=================== Customising Roundup =================== :Version: $Revision: 1.104 $ .. This document borrows from the ZopeBook section on ZPT. The original is at: http://www.zope.org/Documentation/Books/ZopeBook/current/ZPT.stx .. contents:: :depth: 1 What You Can Do =============== Before you get too far, it's probably worth having a quick read of the Roundup `design documentation`_. Customisation of Roundup can take one of six forms: 1. `tracker configuration`_ file changes 2. database, or `tracker schema`_ changes 3. "definition" class `database content`_ changes 4. behavioural changes, through detectors_ 5. `access controls`_ 6. change the `web interface`_ The third case is special because it takes two distinctly different forms depending upon whether the tracker has been initialised or not. The other two may be done at any time, before or after tracker initialisation. Yes, this includes adding or removing properties from classes. Trackers in a Nutshell ====================== Trackers have the following structure: =================== ======================================================== Tracker File Description =================== ======================================================== config.py Holds the basic `tracker configuration`_ dbinit.py Holds the `tracker schema`_ interfaces.py Defines the Web and E-Mail interfaces for the tracker select_db.py Selects the database back-end for the tracker db/ Holds the tracker's database db/files/ Holds the tracker's upload files and messages detectors/ Auditors and reactors for this tracker html/ Web interface templates, images and style sheets =================== ======================================================== Tracker Configuration ===================== The ``config.py`` located in your tracker home contains the basic configuration for the web and e-mail components of roundup's interfaces. As the name suggests, this file is a Python module. This means that any valid python expression may be used in the file. Mostly though, you'll be setting the configuration variables to string values. Python string values must be quoted with either single or double quotes:: 'this is a string' "this is also a string - use it when the value has 'single quotes'" this is not a string - it's not quoted Python strings may use formatting that's almost identical to C string formatting. The ``%`` operator is used to perform the formatting, like so:: 'roundup-admin@%s'%MAIL_DOMAIN this will create a string ``'roundup-admin@tracker.domain.example'`` if MAIL_DOMAIN is set to ``'tracker.domain.example'``. You'll also note some values are set to:: os.path.join(TRACKER_HOME, 'db') or similar. This creates a new string which holds the path to the ``'db'`` directory in the TRACKER_HOME directory. This is just a convenience so if the TRACKER_HOME changes you don't have to edit multiple valoues. The configuration variables available are: **TRACKER_HOME** - ``os.path.split(__file__)[0]`` The tracker home directory. The above default code will automatically determine the tracker home for you, so you can just leave it alone. **MAILHOST** - ``'localhost'`` The SMTP mail host that roundup will use to send e-mail. **MAILUSER** - ``()`` If your SMTP mail host requires a username and password for access, then specify them here. eg. ``MAILUSER = ('username', 'password')`` **MAILHOST_TLS** - ``'no'`` If your SMTP mail host provides or requires TLS (Transport Layer Security) then set ``MAILHOST_TLS = 'yes'`` **MAILHOST_TLS_KEYFILE** - ``''`` If you're using TLS, you may also set MAILHOST_TLS_KEYFILE to the name of a PEM formatted file that contains your private key. **MAILHOST_TLS_CERTFILE** - ``''`` If you're using TLS and have specified a MAILHOST_TLS_KEYFILE, you may also set MAILHOST_TLS_CERTFILE to the name of a PEM formatted certificate chain file. **MAIL_DOMAIN** - ``'tracker.domain.example'`` The domain name used for email addresses. **DATABASE** - ``os.path.join(TRACKER_HOME, 'db')`` This is the directory that the database is going to be stored in. By default it is in the tracker home. **TEMPLATES** - ``os.path.join(TRACKER_HOME, 'html')`` This is the directory that the HTML templates reside in. By default they are in the tracker home. **TRACKER_NAME** - ``'Roundup issue tracker'`` A descriptive name for your roundup tracker. This is sent out in e-mails and appears in the heading of CGI pages. **TRACKER_EMAIL** - ``'issue_tracker@%s'%MAIL_DOMAIN`` The email address that e-mail sent to roundup should go to. Think of it as the tracker's personal e-mail address. **TRACKER_WEB** - ``'http://tracker.example/cgi-bin/roundup.cgi/bugs/'`` The web address that the tracker is viewable at. This will be included in information sent to users of the tracker. The URL **must** include the cgi-bin part or anything else that is required to get to the home page of the tracker. You **must** include a trailing '/' in the URL. **ADMIN_EMAIL** - ``'roundup-admin@%s'%MAIL_DOMAIN`` The email address that roundup will complain to if it runs into trouble. **EMAIL_FROM_TAG** - ``''`` Additional text to include in the "name" part of the ``From:`` address used in nosy messages. If the sending user is "Foo Bar", the ``From:`` line is usually:: "Foo Bar" The EMAIL_FROM_TAG goes inside the "Foo Bar" quotes like so:: "Foo Bar EMAIL_FROM_TAG" **MESSAGES_TO_AUTHOR** - ``'new'``, ``'yes'`` or``'no'`` Send nosy messages to the author of the message? If 'new' is used, then the author will only be sent the message when the message creates a new issue. If 'yes' then the author will always be sent a copy of the message they wrote. **ADD_AUTHOR_TO_NOSY** - ``'new'``, ``'yes'`` or ``'no'`` Does the author of a message get placed on the nosy list automatically? If ``'new'`` is used, then the author will only be added when a message creates a new issue. If ``'yes'``, then the author will be added on followups too. If ``'no'``, they're never added to the nosy. **ADD_RECIPIENTS_TO_NOSY** - ``'new'``, ``'yes'`` or ``'no'`` Do the recipients (To:, Cc:) of a message get placed on the nosy list? If ``'new'`` is used, then the recipients will only be added when a message creates a new issue. If ``'yes'``, then the recipients will be added on followups too. If ``'no'``, they're never added to the nosy. **EMAIL_SIGNATURE_POSITION** - ``'top'``, ``'bottom'`` or ``'none'`` Where to place the email signature in messages that Roundup generates. **EMAIL_KEEP_QUOTED_TEXT** - ``'yes'`` or ``'no'`` Keep email citations. Citations are the part of e-mail which the sender has quoted in their reply to previous e-mail. **EMAIL_LEAVE_BODY_UNCHANGED** - ``'no'`` Preserve the email body as is. Enabiling this will cause the entire message body to be stored, including all citations and signatures. It should be either ``'yes'`` or ``'no'``. **MAIL_DEFAULT_CLASS** - ``'issue'`` or ``''`` Default class to use in the mailgw if one isn't supplied in email subjects. To disable, comment out the variable below or leave it blank. The default config.py is given below - as you can see, the MAIL_DOMAIN must be edited before any interaction with the tracker is attempted.:: # roundup home is this package's directory TRACKER_HOME=os.path.split(__file__)[0] # The SMTP mail host that roundup will use to send mail MAILHOST = 'localhost' # The domain name used for email addresses. MAIL_DOMAIN = 'your.tracker.email.domain.example' # This is the directory that the database is going to be stored in DATABASE = os.path.join(TRACKER_HOME, 'db') # This is the directory that the HTML templates reside in TEMPLATES = os.path.join(TRACKER_HOME, 'html') # A descriptive name for your roundup tracker TRACKER_NAME = 'Roundup issue tracker' # The email address that mail to roundup should go to TRACKER_EMAIL = 'issue_tracker@%s'%MAIL_DOMAIN # The web address that the tracker is viewable at. This will be # included in information sent to users of the tracker. The URL MUST # include the cgi-bin part or anything else that is required to get # to the home page of the tracker. You MUST include a trailing '/' # in the URL. TRACKER_WEB = 'http://tracker.example/cgi-bin/roundup.cgi/bugs/' # The email address that roundup will complain to if it runs into # trouble ADMIN_EMAIL = 'roundup-admin@%s'%MAIL_DOMAIN # Additional text to include in the "name" part of the From: address # used in nosy messages. If the sending user is "Foo Bar", the From: # line is usually: "Foo Bar" # the EMAIL_FROM_TAG goes inside the "Foo Bar" quotes like so: # "Foo Bar EMAIL_FROM_TAG" EMAIL_FROM_TAG = "" # Send nosy messages to the author of the message MESSAGES_TO_AUTHOR = 'no' # either 'yes' or 'no' # Does the author of a message get placed on the nosy list # automatically? If 'new' is used, then the author will only be # added when a message creates a new issue. If 'yes', then the # author will be added on followups too. If 'no', they're never # added to the nosy. ADD_AUTHOR_TO_NOSY = 'new' # one of 'yes', 'no', 'new' # Do the recipients (To:, Cc:) of a message get placed on the nosy # list? If 'new' is used, then the recipients will only be added # when a message creates a new issue. If 'yes', then the recipients # will be added on followups too. If 'no', they're never added to # the nosy. ADD_RECIPIENTS_TO_NOSY = 'new' # either 'yes', 'no', 'new' # Where to place the email signature EMAIL_SIGNATURE_POSITION = 'bottom' # one of 'top', 'bottom', 'none' # Keep email citations EMAIL_KEEP_QUOTED_TEXT = 'no' # either 'yes' or 'no' # Preserve the email body as is EMAIL_LEAVE_BODY_UNCHANGED = 'no' # either 'yes' or 'no' # Default class to use in the mailgw if one isn't supplied in email # subjects. To disable, comment out the variable below or leave it # blank. Examples: MAIL_DEFAULT_CLASS = 'issue' # use "issue" class by default #MAIL_DEFAULT_CLASS = '' # disable (or just comment the var out) # # SECURITY DEFINITIONS # # define the Roles that a user gets when they register with the # tracker these are a comma-separated string of role names (e.g. # 'Admin,User') NEW_WEB_USER_ROLES = 'User' NEW_EMAIL_USER_ROLES = 'User' Tracker Schema ============== Note: if you modify the schema, you'll most likely need to edit the `web interface`_ HTML template files and `detectors`_ to reflect your changes. A tracker schema defines what data is stored in the tracker's database. Schemas are defined using Python code in the ``dbinit.py`` module of your tracker. The "classic" schema looks like this (see below for the meaning of ``'setkey'``):: pri = Class(db, "priority", name=String(), order=String()) pri.setkey("name") stat = Class(db, "status", name=String(), order=String()) stat.setkey("name") keyword = Class(db, "keyword", name=String()) keyword.setkey("name") user = Class(db, "user", username=String(), organisation=String(), password=String(), address=String(), realname=String(), phone=String()) user.setkey("username") msg = FileClass(db, "msg", author=Link("user"), summary=String(), date=Date(), recipients=Multilink("user"), files=Multilink("file")) file = FileClass(db, "file", name=String(), type=String()) issue = IssueClass(db, "issue", topic=Multilink("keyword"), status=Link("status"), assignedto=Link("user"), priority=Link("priority")) issue.setkey('title') What you can't do to the schema ------------------------------- You must never: **Remove the users class** This class is the only *required* class in Roundup. Similarly, its username, password and address properties must never be removed. **Change the type of a property** Property types must *never* be changed - the database simply doesn't take this kind of action into account. Note that you can't just remove a property and re-add it as a new type either. If you wanted to make the assignedto property a Multilink, you'd need to create a new property assignedto_list and remove the old assignedto property. What you can do to the schema ----------------------------- Your schema may be changed at any time before or after the tracker has been initialised (or used). You may: **Add new properties to classes, or add whole new classes** This is painless and easy to do - there are generally no repurcussions from adding new information to a tracker's schema. **Remove properties** Removing properties is a little more tricky - you need to make sure that the property is no longer used in the `web interface`_ *or* by the detectors_. Classes and Properties - creating a new information store --------------------------------------------------------- In the tracker above, we've defined 7 classes of information: priority Defines the possible levels of urgency for issues. status Defines the possible states of processing the issue may be in. keyword Initially empty, will hold keywords useful for searching issues. user Initially holding the "admin" user, will eventually have an entry for all users using roundup. msg Initially empty, will hold all e-mail messages sent to or generated by roundup. file Initially empty, will hold all files attached to issues. issue Initially empty, this is where the issue information is stored. We define the "priority" and "status" classes to allow two things: reduction in the amount of information stored on the issue and more powerful, accurate searching of issues by priority and status. By only requiring a link on the issue (which is stored as a single number) we reduce the chance that someone mis-types a priority or status - or simply makes a new one up. Class and Items ~~~~~~~~~~~~~~~ A Class defines a particular class (or type) of data that will be stored in the database. A class comprises one or more properties, which gives the information about the class items. The actual data entered into the database, using ``class.create()``, are called items. They have a special immutable property called ``'id'``. We sometimes refer to this as the *itemid*. Properties ~~~~~~~~~~ A Class is comprised of one or more properties of the following types: * String properties are for storing arbitrary-length strings. * Password properties are for storing encoded arbitrary-length strings. The default encoding is defined on the ``roundup.password.Password`` class. * Date properties store date-and-time stamps. Their values are Timestamp objects. * Number properties store numeric values. * Boolean properties store on/off, yes/no, true/false values. * A Link property refers to a single other item selected from a specified class. The class is part of the property; the value is an integer, the id of the chosen item. * A Multilink property refers to possibly many items in a specified class. The value is a list of integers. FileClass ~~~~~~~~~ FileClasses save their "content" attribute off in a separate file from the rest of the database. This reduces the number of large entries in the database, which generally makes databases more efficient, and also allows us to use command-line tools to operate on the files. They are stored in the files sub-directory of the ``'db'`` directory in your tracker. IssueClass ~~~~~~~~~~ IssueClasses automatically include the "messages", "files", "nosy", and "superseder" properties. The messages and files properties list the links to the messages and files related to the issue. The nosy property is a list of links to users who wish to be informed of changes to the issue - they get "CC'ed" e-mails when messages are sent to or generated by the issue. The nosy reactor (in the ``'detectors'`` directory) handles this action. The superseder link indicates an issue which has superseded this one. They also have the dynamically generated "creation", "activity" and "creator" properties. The value of the "creation" property is the date when an item was created, and the value of the "activity" property is the date when any property on the item was last edited (equivalently, these are the dates on the first and last records in the item's journal). The "creator" property holds a link to the user that created the issue. setkey(property) ~~~~~~~~~~~~~~~~ Select a String property of the class to be the key property. The key property must be unique, and allows references to the items in the class by the content of the key property. That is, we can refer to users by their username: for example, let's say that there's an issue in roundup, issue 23. There's also a user, richard, who happens to be user 2. To assign an issue to him, we could do either of:: roundup-admin set issue23 assignedto=2 or:: roundup-admin set issue23 assignedto=richard Note, the same thing can be done in the web and e-mail interfaces. If a class does not have an "order" property, the key is also used to sort instances of the class when it is rendered in the user interface. (If a class has no "order" property, sorting is by the labelproperty of the class. This is computed, in order of precedence, as the key, the "name", the "title", or the first property alphabetically.) create(information) ~~~~~~~~~~~~~~~~~~~ Create an item in the database. This is generally used to create items in the "definitional" classes like "priority" and "status". Examples of adding to your schema --------------------------------- TODO Detectors - adding behaviour to your tracker ============================================ .. _detectors: Detectors are initialised every time you open your tracker database, so you're free to add and remove them any time, even after the database is initialised via the "roundup-admin initialise" command. The detectors in your tracker fire *before* (**auditors**) and *after* (**reactors**) changes to the contents of your database. They are Python modules that sit in your tracker's ``detectors`` directory. You will have some installed by default - have a look. You can write new detectors or modify the existing ones. The existing detectors installed for you are: **nosyreaction.py** This provides the automatic nosy list maintenance and email sending. The nosy reactor (``nosyreaction``) fires when new messages are added to issues. The nosy auditor (``updatenosy``) fires when issues are changed, and figures out what changes need to be made to the nosy list (such as adding new authors, etc.) **statusauditor.py** This provides the ``chatty`` auditor which changes the issue status from ``unread`` or ``closed`` to ``chatting`` if new messages appear. It also provides the ``presetunread`` auditor which pre-sets the status to ``unread`` on new items if the status isn't explicitly defined. **messagesummary.py** Generates the ``summary`` property for new messages based on the message content. **userauditor.py** Verifies the content of some of the user fields (email addresses and roles lists). If you don't want this default behaviour, you're completely free to change or remove these detectors. See the detectors section in the `design document`__ for details of the interface for detectors. __ design.html Sample additional detectors that have been found useful will appear in the ``'detectors'`` directory of the Roundup distribution. If you want to use one, copy it to the ``'detectors'`` of your tracker instance: **newissuecopy.py** This detector sends an email to a team address whenever a new issue is created. The address is hard-coded into the detector, so edit it before you use it (look for the text 'team@team.host') or you'll get email errors! The detector code:: from roundup import roundupdb def newissuecopy(db, cl, nodeid, oldvalues): ''' Copy a message about new issues to a team address. ''' # so use all the messages in the create change_note = cl.generateCreateNote(nodeid) # send a copy to the nosy list for msgid in cl.get(nodeid, 'messages'): try: # note: last arg must be a list cl.send_message(nodeid, msgid, change_note, ['team@team.host']) except roundupdb.MessageSendError, message: raise roundupdb.DetectorError, message def init(db): db.issue.react('create', newissuecopy) Database Content ================ Note: if you modify the content of definitional classes, you'll most likely need to edit the tracker `detectors`_ to reflect your changes. Customisation of the special "definitional" classes (eg. status, priority, resolution, ...) may be done either before or after the tracker is initialised. The actual method of doing so is completely different in each case though, so be careful to use the right one. **Changing content before tracker initialisation** Edit the dbinit module in your tracker to alter the items created in using the ``create()`` methods. **Changing content after tracker initialisation** As the "admin" user, click on the "class list" link in the web interface to bring up a list of all database classes. Click on the name of the class you wish to change the content of. You may also use the ``roundup-admin`` interface's create, set and retire methods to add, alter or remove items from the classes in question. See "`adding a new field to the classic schema`_" for an example that requires database content changes. Access Controls =============== A set of Permissions is built into the security module by default: - Edit (everything) - View (everything) The default interfaces define: - Web Registration - Web Access - Web Roles - Email Registration - Email Access These are hooked into the default Roles: - Admin (Edit everything, View everything, Web Roles) - User (Web Access, Email Access) - Anonymous (Web Registration, Email Registration) And finally, the "admin" user gets the "Admin" Role, and the "anonymous" user gets "Anonymous" assigned when the database is initialised on installation. The two default schemas then define: - Edit issue, View issue (both) - Edit file, View file (both) - Edit msg, View msg (both) - Edit support, View support (extended only) and assign those Permissions to the "User" Role. Put together, these settings appear in the ``open()`` function of the tracker ``dbinit.py`` (the following is taken from the "minimal" template's ``dbinit.py``):: # # SECURITY SETTINGS # # new permissions for this schema for cl in ('user', ): db.security.addPermission(name="Edit", klass=cl, description="User is allowed to edit "+cl) db.security.addPermission(name="View", klass=cl, description="User is allowed to access "+cl) # and give the regular users access to the web and email interface p = db.security.getPermission('Web Access') db.security.addPermissionToRole('User', p) p = db.security.getPermission('Email Access') db.security.addPermissionToRole('User', p) # May users view other user information? Comment these lines out # if you don't want them to p = db.security.getPermission('View', 'user') db.security.addPermissionToRole('User', p) # Assign the appropriate permissions to the anonymous user's # Anonymous role. Choices here are: # - Allow anonymous users to register through the web p = db.security.getPermission('Web Registration') db.security.addPermissionToRole('Anonymous', p) # - Allow anonymous (new) users to register through the email # gateway p = db.security.getPermission('Email Registration') db.security.addPermissionToRole('Anonymous', p) New User Roles -------------- New users are assigned the Roles defined in the config file as: - NEW_WEB_USER_ROLES - NEW_EMAIL_USER_ROLES Changing Access Controls ------------------------ You may alter the configuration variables to change the Role that new web or email users get, for example to not give them access to the web interface if they register through email. You may use the ``roundup-admin`` "``security``" command to display the current Role and Permission configuration in your tracker. Adding a new Permission ~~~~~~~~~~~~~~~~~~~~~~~ When adding a new Permission, you will need to: 1. add it to your tracker's dbinit so it is created 2. enable it for the Roles that should have it (verify with "``roundup-admin security``") 3. add it to the relevant HTML interface templates 4. add it to the appropriate xxxPermission methods on in your tracker interfaces module Example Scenarios ~~~~~~~~~~~~~~~~~ **automatic registration of users in the e-mail gateway** By giving the "anonymous" user the "Email Registration" Role, any unidentified user will automatically be registered with the tracker (with no password, so they won't be able to log in through the web until an admin sets their password). Note: this is the default behaviour in the tracker templates that ship with Roundup. **anonymous access through the e-mail gateway** Give the "anonymous" user the "Email Access" and ("Edit", "issue") Roles but do not not give them the "Email Registration" Role. This means that when an unknown user sends email into the tracker, they're automatically logged in as "anonymous". Since they don't have the "Email Registration" Role, they won't be automatically registered, but since "anonymous" has permission to use the gateway, they'll still be able to submit issues. Note that the Sender information - their email address - will not be available - they're *anonymous*. **only developers may be assigned issues** Create a new Permission called "Fixer" for the "issue" class. Create a new Role "Developer" which has that Permission, and assign that to the appropriate users. Filter the list of users available in the assignedto list to include only those users. Enforce the Permission with an auditor. See the example `restricting the list of users that are assignable to a task`_. **only managers may sign off issues as complete** Create a new Permission called "Closer" for the "issue" class. Create a new Role "Manager" which has that Permission, and assign that to the appropriate users. In your web interface, only display the "resolved" issue state option when the user has the "Closer" Permissions. Enforce the Permission with an auditor. This is very similar to the previous example, except that the web interface check would look like:: **don't give web access to users who register through email** Create a new Role called "Email User" which has all the Permissions of the normal "User" Role minus the "Web Access" Permission. This will allow users to send in emails to the tracker, but not access the web interface. **let some users edit the details of all users** Create a new Role called "User Admin" which has the Permission for editing users:: db.security.addRole(name='User Admin', description='Managing users') p = db.security.getPermission('Edit', 'user') db.security.addPermissionToRole('User Admin', p) and assign the Role to the users who need the permission. Web Interface ============= .. contents:: :local: :depth: 1 The web interface is provided by the ``roundup.cgi.client`` module and is used by ``roundup.cgi``, ``roundup-server`` and ``ZRoundup`` (``ZRoundup`` is broken, until further notice). In all cases, we determine which tracker is being accessed (the first part of the URL path inside the scope of the CGI handler) and pass control on to the tracker ``interfaces.Client`` class - which uses the ``Client`` class from ``roundup.cgi.client`` - which handles the rest of the access through its ``main()`` method. This means that you can do pretty much anything you want as a web interface to your tracker. Repercussions of changing the tracker schema --------------------------------------------- If you choose to change the `tracker schema`_ you will need to ensure the web interface knows about it: 1. Index, item and search pages for the relevant classes may need to have properties added or removed, 2. The "page" template may require links to be changed, as might the "home" page's content arguments. How requests are processed -------------------------- The basic processing of a web request proceeds as follows: 1. figure out who we are, defaulting to the "anonymous" user 2. figure out what the request is for - we call this the "context" 3. handle any requested action (item edit, search, ...) 4. render the template requested by the context, resulting in HTML output In some situations, exceptions occur: - HTTP Redirect (generally raised by an action) - SendFile (generally raised by ``determine_context``) here we serve up a FileClass "content" property - SendStaticFile (generally raised by ``determine_context``) here we serve up a file from the tracker "html" directory - Unauthorised (generally raised by an action) here the action is cancelled, the request is rendered and an error message is displayed indicating that permission was not granted for the action to take place - NotFound (raised wherever it needs to be) this exception percolates up to the CGI interface that called the client Determining web context ----------------------- To determine the "context" of a request, we look at the URL and the special request variable ``@template``. The URL path after the tracker identifier is examined. Typical URL paths look like: 1. ``/tracker/issue`` 2. ``/tracker/issue1`` 3. ``/tracker/_file/style.css`` 4. ``/cgi-bin/roundup.cgi/tracker/file1`` 5. ``/cgi-bin/roundup.cgi/tracker/file1/kitten.png`` where the "tracker identifier" is "tracker" in the above cases. That means we're looking at "issue", "issue1", "_file/style.css", "file1" and "file1/kitten.png" in the cases above. The path is generally only one entry long - longer paths are handled differently. a. if there is no path, then we are in the "home" context. b. if the path starts with "_file" (as in example 3, "/tracker/_file/style.css"), then the additional path entry, "style.css" specifies the filename of a static file we're to serve up from the tracker "html" directory. Raises a SendStaticFile exception. c. if there is something in the path (as in example 1, "issue"), it identifies the tracker class we're to display. d. if the path is an item designator (as in examples 2 and 4, "issue1" and "file1"), then we're to display a specific item. e. if the path starts with an item designator and is longer than one entry (as in example 5, "file1/kitten.png"), then we're assumed to be handling an item of a ``FileClass``, and the extra path information gives the filename that the client is going to label the download with (i.e. "file1/kitten.png" is nicer to download than "file1"). This raises a ``SendFile`` exception. Both b. and e. stop before we bother to determine the template we're going to use. That's because they don't actually use templates. The template used is specified by the ``@template`` CGI variable, which defaults to: - only classname suplied: "index" - full item designator supplied: "item" Performing actions in web requests ---------------------------------- When a user requests a web page, they may optionally also request for an action to take place. As described in `how requests are processed`_, the action is performed before the requested page is generated. Actions are triggered by using a ``@action`` CGI variable, where the value is one of: **login** Attempt to log a user in. **logout** Log the user out - make them "anonymous". **register** Attempt to create a new user based on the contents of the form and then log them in. **edit** Perform an edit of an item in the database. There are some `special form variables`_ you may use. **new** Add a new item to the database. You may use the same `special form variables`_ as in the "edit" action. **retire** Retire the item in the database. **editCSV** Performs an edit of all of a class' items in one go. See also the *class*.csv templating method which generates the CSV data to be edited, and the ``'_generic.index'`` template which uses both of these features. **search** Mangle some of the form variables: - Set the form ":filter" variable based on the values of the filter variables - if they're set to anything other than "dontcare" then add them to :filter. - Also handle the ":queryname" variable and save off the query to the user's query list. Each of the actions is implemented by a corresponding ``*actionAction*`` (where "action" is the name of the action) method on the ``roundup.cgi.Client`` class, which also happens to be available in your tracker instance as ``interfaces.Client``. So if you need to define new actions, you may add them there (see `defining new web actions`_). Each action also has a corresponding ``*actionPermission*`` (where "action" is the name of the action) method which determines whether the action is permissible given the current user. The base permission checks are: **login** Determine whether the user has permission to log in. Base behaviour is to check the user has "Web Access". **logout** No permission checks are made. **register** Determine whether the user has permission to register. Base behaviour is to check the user has the "Web Registration" Permission. **edit** Determine whether the user has permission to edit this item. Base behaviour is to check whether the user can edit this class. If we're editing the "user" class, users are allowed to edit their own details - unless they try to edit the "roles" property, which requires the special Permission "Web Roles". **new** Determine whether the user has permission to create (or edit) this item. Base behaviour is to check the user can edit this class. No additional property checks are made. Additionally, new user items may be created if the user has the "Web Registration" Permission. **editCSV** Determine whether the user has permission to edit this class. Base behaviour is to check whether the user may edit this class. **search** Determine whether the user has permission to search this class. Base behaviour is to check whether the user may view this class. Special form variables ---------------------- Item properties and their values are edited with html FORM variables and their values. You can: - Change the value of some property of the current item. - Create a new item of any class, and edit the new item's properties, - Attach newly created items to a multilink property of the current item. - Remove items from a multilink property of the current item. - Specify that some properties are required for the edit operation to be successful. In the following, values are variable, "@" may be either ":" or "@", and other text "required" is fixed. Most properties are specified as form variables: ```` property on the current context item ``"@"`` property on the indicated item (for editing related information) Designators name a specific item of a class. ```` Name an existing item of class . ``"-"`` Name the th new item of class . If the form submission is successful, a new item of is created. Within the submitted form, a particular designator of this form always refers to the same new item. Once we have determined the "propname", we look at it to see if it's special: ``@required`` The associated form value is a comma-separated list of property names that must be specified when the form is submitted for the edit operation to succeed. When the is missing, the properties are for the current context item. When is present, they are for the item specified by . The "@required" specifier must come before any of the properties it refers to are assigned in the form. ``@remove@=id(s)`` or ``@add@=id(s)`` The "@add@" and "@remove@" edit actions apply only to Multilink properties. The form value must be a comma-separate list of keys for the class specified by the simple form variable. The listed items are added to (respectively, removed from) the specified property. ``@link@=`` If the edit action is "@link@", the simple form variable must specify a Link or Multilink property. The form value is a comma-separated list of designators. The item corresponding to each designator is linked to the property given by simple form variable. None of the above (ie. just a simple form value) The value of the form variable is converted appropriately, depending on the type of the property. For a Link('klass') property, the form value is a single key for 'klass', where the key field is specified in dbinit.py. For a Multilink('klass') property, the form value is a comma-separated list of keys for 'klass', where the key field is specified in dbinit.py. Note that for simple-form-variables specifiying Link and Multilink properties, the linked-to class must have a key field. For a String() property specifying a filename, the file named by the form value is uploaded. This means we try to set additional properties "filename" and "type" (if they are valid for the class). Otherwise, the property is set to the form value. For Date(), Interval(), Boolean(), and Number() properties, the form value is converted to the appropriate Any of the form variables may be prefixed with a classname or designator. Two special form values are supported for backwards compatibility: @note This is equivalent to:: @link@messages=msg-1 @msg-1@content=value except that in addition, the "author" and "date" properties of "msg-1" are set to the userid of the submitter, and the current time, respectively. @file This is equivalent to:: @link@files=file-1 @file-1@content=value The String content value is handled as described above for file uploads. If both the "@note" and "@file" form variables are specified, the action:: @link@msg-1@files=file-1 is also performed. We also check that FileClass items have a "content" property with actual content, otherwise we remove them from all_props before returning. Default templates ----------------- Most customisation of the web view can be done by modifying the templates in the tracker ``'html'`` directory. There are several types of files in there. The *minimal* template includes: **page.html** This template usually defines the overall look of your tracker. When you view an issue, it appears inside this template. When you view an index, it also appears inside this template. This template defines a macro called "icing" which is used by almost all other templates as a coating for their content, using its "content" slot. It also defines the "head_title" and "body_title" slots to allow setting of the page title. **home.html** the default page displayed when no other page is indicated by the user **home.classlist.html** a special version of the default page that lists the classes in the tracker **classname.item.html** displays an item of the *classname* class **classname.index.html** displays a list of *classname* items **classname.search.html** displays a search page for *classname* items **_generic.index.html** used to display a list of items where there is no ``*classname*.index`` available **_generic.help.html** used to display a "class help" page where there is no ``*classname*.help`` **user.register.html** a special page just for the user class, that renders the registration page **style.css.html** a static file that is served up as-is The *classic* template has a number of additional templates. Note: Remember that you can create any template extension you want to, so if you just want to play around with the templating for new issues, you can copy the current "issue.item" template to "issue.test", and then access the test template using the "@template" URL argument:: http://your.tracker.example/tracker/issue?@template=test and it won't affect your users using the "issue.item" template. How the templates work ---------------------- Basic Templating Actions ~~~~~~~~~~~~~~~~~~~~~~~~ Roundup's templates consist of special attributes on the HTML tags. These attributes form the Template Attribute Language, or TAL. The basic TAL commands are: **tal:define="variable expression; variable expression; ..."** Define a new variable that is local to this tag and its contents. For example:: In this example, the variable "title" is defined as the result of the expression "request/description". The "tal:content" command inside the tag may then use the "title" variable. **tal:condition="expression"** Only keep this tag and its contents if the expression is true. For example::

Display some issue information.

In the example, the

tag and its contents are only displayed if the user has the "View" permission for issues. We consider the number zero, a blank string, an empty list, and the built-in variable nothing to be false values. Nearly every other value is true, including non-zero numbers, and strings with anything in them (even spaces!). **tal:repeat="variable expression"** Repeat this tag and its contents for each element of the sequence that the expression returns, defining a new local variable and a special "repeat" variable for each element. For example:: The example would iterate over the sequence of users returned by "user/list" and define the local variable "u" for each entry. **tal:replace="expression"** Replace this tag with the result of the expression. For example:: The example would replace the tag and its contents with the user's realname. If the user's realname was "Bruce", then the resultant output would be "Bruce". **tal:content="expression"** Replace the contents of this tag with the result of the expression. For example:: user's name appears here The example would replace the contents of the tag with the user's realname. If the user's realname was "Bruce" then the resultant output would be "Bruce". **tal:attributes="attribute expression; attribute expression; ..."** Set attributes on this tag to the results of expressions. For example:: My Details In the example, the "href" attribute of the tag is set to the value of the "string:user${request/user/id}" expression, which will be something like "user123". **tal:omit-tag="expression"** Remove this tag (but not its contents) if the expression is true. For example:: Hello, world! would result in output of:: Hello, world! Note that the commands on a given tag are evaulated in the order above, so *define* comes before *condition*, and so on. Additionally, you may include tags such as , which are removed from output. Its content is kept, but the tag itself is not (so don't go using any "tal:attributes" commands on it). This is useful for making arbitrary blocks of HTML conditional or repeatable (very handy for repeating multiple table rows, which would othewise require an illegal tag placement to effect the repeat). Templating Expressions ~~~~~~~~~~~~~~~~~~~~~~ The expressions you may use in the attribute values may be one of the following forms: **Path Expressions** - eg. ``item/status/checklist`` These are object attribute / item accesses. Roughly speaking, the path ``item/status/checklist`` is broken into parts ``item``, ``status`` and ``checklist``. The ``item`` part is the root of the expression. We then look for a ``status`` attribute on ``item``, or failing that, a ``status`` item (as in ``item['status']``). If that fails, the path expression fails. When we get to the end, the object we're left with is evaluated to get a string - if it is a method, it is called; if it is an object, it is stringified. Path expressions may have an optional ``path:`` prefix, but they are the default expression type, so it's not necessary. If an expression evaluates to ``default``, then the expression is "cancelled" - whatever HTML already exists in the template will remain (tag content in the case of ``tal:content``, attributes in the case of ``tal:attributes``). If an expression evaluates to ``nothing`` then the target of the expression is removed (tag content in the case of ``tal:content``, attributes in the case of ``tal:attributes`` and the tag itself in the case of ``tal:replace``). If an element in the path may not exist, then you can use the ``|`` operator in the expression to provide an alternative. So, the expression ``request/form/foo/value | default`` would simply leave the current HTML in place if the "foo" form variable doesn't exist. You may use the python function ``path``, as in ``path("item/status")``, to embed path expressions in Python expressions. **String Expressions** - eg. ``string:hello ${user/name}`` These expressions are simple string interpolations - though they can be just plain strings with no interpolation if you want. The expression in the ``${ ... }`` is just a path expression as above. **Python Expressions** - eg. ``python: 1+1`` These expressions give the full power of Python. All the "root level" variables are available, so ``python:item.status.checklist()`` would be equivalent to ``item/status/checklist``, assuming that ``checklist`` is a method. Modifiers: **structure** - eg. ``structure python:msg.content.plain(hyperlink=1)`` The result of expressions are normally *escaped* to be safe for HTML display (all "<", ">" and "&" are turned into special entities). The ``structure`` expression modifier turns off this escaping - the result of the expression is now assumed to be HTML, which is passed to the web browser for rendering. **not:** - eg. ``not:python:1=1`` This simply inverts the logical true/false value of another expression. Template Macros ~~~~~~~~~~~~~~~ Macros are used in Roundup to save us from repeating the same common page stuctures over and over. The most common (and probably only) macro you'll use is the "icing" macro defined in the "page" template. Macros are generated and used inside your templates using special attributes similar to the `basic templating actions`_. In this case, though, the attributes belong to the Macro Expansion Template Attribute Language, or METAL. The macro commands are: **metal:define-macro="macro name"** Define that the tag and its contents are now a macro that may be inserted into other templates using the *use-macro* command. For example:: ... defines a macro called "page" using the ```` tag and its contents. Once defined, macros are stored on the template they're defined on in the ``macros`` attribute. You can access them later on through the ``templates`` variable, eg. the most common ``templates/page/macros/icing`` to access the "page" macro of the "page" template. **metal:use-macro="path expression"** Use a macro, which is identified by the path expression (see above). This will replace the current tag with the identified macro contents. For example:: ... will replace the tag and its contents with the "page" macro of the "page" template. **metal:define-slot="slot name"** and **metal:fill-slot="slot name"** To define *dynamic* parts of the macro, you define "slots" which may be filled when the macro is used with a *use-macro* command. For example, the ``templates/page/macros/icing`` macro defines a slot like so:: title goes here In your *use-macro* command, you may now use a *fill-slot* command like this:: My Title where the tag that fills the slot completely replaces the one defined as the slot in the macro. Note that you may not mix METAL and TAL commands on the same tag, but TAL commands may be used freely inside METAL-using tags (so your *fill-slots* tags may have all manner of TAL inside them). Information available to templates ---------------------------------- Note: this is implemented by ``roundup.cgi.templating.RoundupPageTemplate`` The following variables are available to templates. **context** The current context. This is either None, a `hyperdb class wrapper`_ or a `hyperdb item wrapper`_ **request** Includes information about the current request, including: - the current index information (``filterspec``, ``filter`` args, ``properties``, etc) parsed out of the form. - methods for easy filterspec link generation - *user*, the current user item as an HTMLItem instance - *form* The current CGI form information as a mapping of form argument name to value **config** This variable holds all the values defined in the tracker config.py file (eg. TRACKER_NAME, etc.) **db** The current database, used to access arbitrary database items. **templates** Access to all the tracker templates by name. Used mainly in *use-macro* commands. **utils** This variable makes available some utility functions like batching. **nothing** This is a special variable - if an expression evaluates to this, then the tag (in the case of a ``tal:replace``), its contents (in the case of ``tal:content``) or some attributes (in the case of ``tal:attributes``) will not appear in the the output. So, for example:: Hello, World! would result in:: Hello, World! **default** Also a special variable - if an expression evaluates to this, then the existing HTML in the template will not be replaced or removed, it will remain. So:: Hello, World! would result in:: Hello, World! The context variable ~~~~~~~~~~~~~~~~~~~~ The *context* variable is one of three things based on the current context (see `determining web context`_ for how we figure this out): 1. if we're looking at a "home" page, then it's None 2. if we're looking at a specific hyperdb class, it's a `hyperdb class wrapper`_. 3. if we're looking at a specific hyperdb item, it's a `hyperdb item wrapper`_. If the context is not None, we can access the properties of the class or item. The only real difference between cases 2 and 3 above are: 1. the properties may have a real value behind them, and this will appear if the property is displayed through ``context/property`` or ``context/property/field``. 2. the context's "id" property will be a false value in the second case, but a real, or true value in the third. Thus we can determine whether we're looking at a real item from the hyperdb by testing "context/id". Hyperdb class wrapper ::::::::::::::::::::: Note: this is implemented by the ``roundup.cgi.templating.HTMLClass`` class. This wrapper object provides access to a hyperb class. It is used primarily in both index view and new item views, but it's also usable anywhere else that you wish to access information about a class, or the items of a class, when you don't have a specific item of that class in mind. We allow access to properties. There will be no "id" property. The value accessed through the property will be the current value of the same name from the CGI form. There are several methods available on these wrapper objects: =========== ============================================================= Method Description =========== ============================================================= properties return a `hyperdb property wrapper`_ for all of this class's properties. list lists all of the active (not retired) items in the class. csv return the items of this class as a chunk of CSV text. propnames lists the names of the properties of this class. filter lists of items from this class, filtered and sorted by the current *request* filterspec/filter/sort/group args classhelp display a link to a javascript popup containing this class' "help" template. submit generate a submit button (and action hidden element) renderWith render this class with the given template. history returns 'New node - no history' :) is_edit_ok is the user allowed to Edit the current class? is_view_ok is the user allowed to View the current class? =========== ============================================================= Note that if you have a property of the same name as one of the above methods, you'll need to access it using a python "item access" expression. For example:: python:context['list'] will access the "list" property, rather than the list method. Hyperdb item wrapper :::::::::::::::::::: Note: this is implemented by the ``roundup.cgi.templating.HTMLItem`` class. This wrapper object provides access to a hyperb item. We allow access to properties. There will be no "id" property. The value accessed through the property will be the current value of the same name from the CGI form. There are several methods available on these wrapper objects: =============== ======================================================== Method Description =============== ======================================================== submit generate a submit button (and action hidden element) journal return the journal of the current item (**not implemented**) history render the journal of the current item as HTML renderQueryForm specific to the "query" class - render the search form for the query hasPermission specific to the "user" class - determine whether the user has a Permission is_edit_ok is the user allowed to Edit the current item? is_view_ok is the user allowed to View the current item? =============== ======================================================== Note that if you have a property of the same name as one of the above methods, you'll need to access it using a python "item access" expression. For example:: python:context['journal'] will access the "journal" property, rather than the journal method. Hyperdb property wrapper :::::::::::::::::::::::: Note: this is implemented by subclasses of the ``roundup.cgi.templating.HTMLProperty`` class (``HTMLStringProperty``, ``HTMLNumberProperty``, and so on). This wrapper object provides access to a single property of a class. Its value may be either: 1. if accessed through a `hyperdb item wrapper`_, then it's a value from the hyperdb 2. if access through a `hyperdb class wrapper`_, then it's a value from the CGI form The property wrapper has some useful attributes: =============== ======================================================== Attribute Description =============== ======================================================== _name the name of the property _value the value of the property if any - this is the actual value retrieved from the hyperdb for this property =============== ======================================================== There are several methods available on these wrapper objects: =========== ================================================================ Method Description =========== ================================================================ plain render a "plain" representation of the property. This method may take two arguments: escape If true, escape the text so it is HTML safe (default: no). The reason this defaults to off is that text is usually escaped at a later stage by the TAL commands, unless the "structure" option is used in the template. The following ``tal:content`` expressions are all equivalent:: "structure python:msg.content.plain(escape=1)" "python:msg.content.plain()" "msg/content/plain" "msg/content" Usually you'll only want to use the escape option in a complex expression. hyperlink If true, turn URLs, email addresses and hyperdb item designators in the text into hyperlinks (default: no). Note that you'll need to use the "structure" TAL option if you want to use this ``tal:content`` expression:: "structure python:msg.content.plain(hyperlink=1)" Note also that the text is automatically HTML-escaped before the hyperlinking transformation. hyperlinked The same as msg.content.plain(hyperlink=1), but nicer:: "structure msg/content/hyperlinked" field render an appropriate form edit field for the property - for most types this is a text entry box, but for Booleans it's a tri-state yes/no/neither selection. stext only on String properties - render the value of the property as StructuredText (requires the StructureText module to be installed separately) multiline only on String properties - render a multiline form edit field for the property email only on String properties - render the value of the property as an obscured email address confirm only on Password properties - render a second form edit field for the property, used for confirmation that the user typed the password correctly. Generates a field with name "name:confirm". now only on Date properties - return the current date as a new property reldate only on Date properties - render the interval between the date and now local only on Date properties - return this date as a new property with some timezone offset pretty only on Interval properties - render the interval in a pretty format (eg. "yesterday") menu only on Link and Multilink properties - render a form select list for this property reverse only on Multilink properties - produce a list of the linked items in reverse order =========== ================================================================ The request variable ~~~~~~~~~~~~~~~~~~~~ Note: this is implemented by the ``roundup.cgi.templating.HTMLRequest`` class. The request variable is packed with information about the current request. .. taken from ``roundup.cgi.templating.HTMLRequest`` docstring =========== ============================================================ Variable Holds =========== ============================================================ form the CGI form as a cgi.FieldStorage env the CGI environment variables base the base URL for this tracker user a HTMLUser instance for this user classname the current classname (possibly None) template the current template (suffix, also possibly None) form the current CGI form variables in a FieldStorage =========== ============================================================ **Index page specific variables (indexing arguments)** =========== ============================================================ Variable Holds =========== ============================================================ columns dictionary of the columns to display in an index page show a convenience access to columns - request/show/colname will be true if the columns should be displayed, false otherwise sort index sort column (direction, column name) group index grouping property (direction, column name) filter properties to filter the index on filterspec values to filter the index on search_text text to perform a full-text search on for an index =========== ============================================================ There are several methods available on the request variable: =============== ======================================================== Method Description =============== ======================================================== description render a description of the request - handle for the page title indexargs_form render the current index args as form elements indexargs_url render the current index args as a URL base_javascript render some javascript that is used by other components of the templating batch run the current index args through a filter and return a list of items (see `hyperdb item wrapper`_, and `batching`_) =============== ======================================================== The form variable ::::::::::::::::: The form variable is a bit special because it's actually a python FieldStorage object. That means that you have two ways to access its contents. For example, to look up the CGI form value for the variable "name", use the path expression:: request/form/name/value or the python expression:: python:request.form['name'].value Note the "item" access used in the python case, and also note the explicit "value" attribute we have to access. That's because the form variables are stored as MiniFieldStorages. If there's more than one "name" value in the form, then the above will break since ``request/form/name`` is actually a *list* of MiniFieldStorages. So it's best to know beforehand what you're dealing with. The db variable ~~~~~~~~~~~~~~~ Note: this is implemented by the ``roundup.cgi.templating.HTMLDatabase`` class. Allows access to all hyperdb classes as attributes of this variable. If you want access to the "user" class, for example, you would use:: db/user python:db.user Also, the current id of the current user is available as ``db.getuid()``. This isn't so useful in templates (where you have ``request/user``), but it can be useful in detectors or interfaces. The access results in a `hyperdb class wrapper`_. The templates variable ~~~~~~~~~~~~~~~~~~~~~~ Note: this is implemented by the ``roundup.cgi.templating.Templates`` class. This variable doesn't have any useful methods defined. It supports being used in expressions to access the templates, and consequently the template macros. You may access the templates using the following path expression:: templates/name or the python expression:: templates[name] where "name" is the name of the template you wish to access. The template has one useful attribute, namely "macros". To access a specific macro (called "macro_name"), use the path expression:: templates/name/macros/macro_name or the python expression:: templates[name].macros[macro_name] The utils variable ~~~~~~~~~~~~~~~~~~ Note: this is implemented by the ``roundup.cgi.templating.TemplatingUtils`` class, but it may be extended as described below. =============== ======================================================== Method Description =============== ======================================================== Batch return a batch object using the supplied list =============== ======================================================== You may add additional utility methods by writing them in your tracker ``interfaces.py`` module's ``TemplatingUtils`` class. See `adding a time log to your issues`_ for an example. The TemplatingUtils class itself will have a single attribute, ``client``, which may be used to access the ``client.db`` when you need to perform arbitrary database queries. Batching :::::::: Use Batch to turn a list of items, or item ids of a given class, into a series of batches. Its usage is:: python:utils.Batch(sequence, size, start, end=0, orphan=0, overlap=0) or, to get the current index batch:: request/batch The parameters are: ========= ============================================================== Parameter Usage ========= ============================================================== sequence a list of HTMLItems size how big to make the sequence. start where to start (0-indexed) in the sequence. end where to end (0-indexed) in the sequence. orphan if the next batch would contain less items than this value, then it is combined with this batch overlap the number of items shared between adjacent batches ========= ============================================================== All of the parameters are assigned as attributes on the batch object. In addition, it has several more attributes: =============== ======================================================== Attribute Description =============== ======================================================== start indicates the start index of the batch. *Note: unlike the argument, is a 1-based index (I know, lame)* first indicates the start index of the batch *as a 0-based index* length the actual number of elements in the batch sequence_length the length of the original, unbatched, sequence. =============== ======================================================== And several methods: =============== ======================================================== Method Description =============== ======================================================== previous returns a new Batch with the previous batch settings next returns a new Batch with the next batch settings propchanged detect if the named property changed on the current item when compared to the last item =============== ======================================================== An example of batching::
Existing Keywords
keyword here
... which will produce a table with four columns containing the items of the "keyword" class (well, their "name" anyway). Displaying Properties --------------------- Properties appear in the user interface in three contexts: in indices, in editors, and as search arguments. For each type of property, there are several display possibilities. For example, in an index view, a string property may just be printed as a plain string, but in an editor view, that property may be displayed in an editable field. Index Views ----------- This is one of the class context views. It is also the default view for classes. The template used is "*classname*.index". Index View Specifiers ~~~~~~~~~~~~~~~~~~~~~ An index view specifier (URL fragment) looks like this (whitespace has been added for clarity):: /issue?status=unread,in-progress,resolved& topic=security,ui& :group=+priority& :sort==activity& :filters=status,topic& :columns=title,status,fixer The index view is determined by two parts of the specifier: the layout part and the filter part. The layout part consists of the query parameters that begin with colons, and it determines the way that the properties of selected items are displayed. The filter part consists of all the other query parameters, and it determines the criteria by which items are selected for display. The filter part is interactively manipulated with the form widgets displayed in the filter section. The layout part is interactively manipulated by clicking on the column headings in the table. The filter part selects the union of the sets of items with values matching any specified Link properties and the intersection of the sets of items with values matching any specified Multilink properties. The example specifies an index of "issue" items. Only items with a "status" of either "unread" or "in-progress" or "resolved" are displayed, and only items with "topic" values including both "security" and "ui" are displayed. The items are grouped by priority, arranged in ascending order; and within groups, sorted by activity, arranged in descending order. The filter section shows filters for the "status" and "topic" properties, and the table includes columns for the "title", "status", and "fixer" properties. Searching Views --------------- Note: if you add a new column to the ``:columns`` form variable potentials then you will need to add the column to the appropriate `index views`_ template so that it is actually displayed. This is one of the class context views. The template used is typically "*classname*.search". The form on this page should have "search" as its ``@action`` variable. The "search" action: - sets up additional filtering, as well as performing indexed text searching - sets the ``:filter`` variable correctly - saves the query off if ``:query_name`` is set. The search page should lay out any fields that you wish to allow the user to search on. If your schema contains a large number of properties, you should be wary of making all of those properties available for searching, as this can cause confusion. If the additional properties are Strings, consider having their value indexed, and then they will be searchable using the full text indexed search. This is both faster, and more useful for the end user. The two special form values on search pages which are handled by the "search" action are: :search_text Text with which to perform a search of the text index. Results from that search will be used to limit the results of other filters (using an intersection operation) :query_name If supplied, the search parameters (including :search_text) will be saved off as a the query item and registered against the user's queries property. Note that the *classic* template schema has this ability, but the *minimal* template schema does not. Item Views ---------- The basic view of a hyperdb item is provided by the "*classname*.item" template. It generally has three sections; an "editor", a "spool" and a "history" section. Editor Section ~~~~~~~~~~~~~~ The editor section is used to manipulate the item - it may be a static display if the user doesn't have permission to edit the item. Here's an example of a basic editor template (this is the default "classic" template issue item edit form - from the "issue.item.html" template)::
Title title
Priority priority Status status
Superseder
View: 		Nosy List
Assigned To assignedto menu    
Change Note
File
  submit button will go here
When a change is submitted, the system automatically generates a message describing the changed properties. As shown in the example, the editor template can use the ":note" and ":file" fields, which are added to the standard changenote message generated by Roundup. Form values ::::::::::: We have a number of ways to pull properties out of the form in order to meet the various needs of: 1. editing the current item (perhaps an issue item) 2. editing information related to the current item (eg. messages or attached files) 3. creating new information to be linked to the current item (eg. time spent on an issue) In the following, ```` values are variable, ":" may be one of ":" or "@", and other text ("required") is fixed. Properties are specified as form variables: ```` property on the current context item ``:`` property on the indicated item (for editing related information) ``-:`` property on the Nth new item of classname (generally for creating new items to attach to the current item) Once we have determined the "propname", we check to see if it is one of the special form values: ``@required`` The named property values must be supplied or a ValueError will be raised. ``@remove@=id(s)`` The ids will be removed from the multilink property. ``:add:=id(s)`` The ids will be added to the multilink property. ``:link:=`` Used to add a link to new items created during edit. These are collected and returned in ``all_links``. This will result in an additional linking operation (either Link set or Multilink append) after the edit/create is done using ``all_props`` in ``_editnodes``. The on the current item will be set/appended the id of the newly created item of class (where must be -). Any of the form variables may be prefixed with a classname or designator. Two special form values are supported for backwards compatibility: ``:note`` create a message (with content, author and date), linked to the context item. This is ALWAYS designated "msg-1". ``:file`` create a file, attached to the current item and any message created by :note. This is ALWAYS designated "file-1". Spool Section ~~~~~~~~~~~~~ The spool section lists related information like the messages and files of an issue. TODO History Section ~~~~~~~~~~~~~~~ The final section displayed is the history of the item - its database journal. This is generally generated with the template:: *To be done:* *The actual history entries of the item may be accessed for manual templating through the "journal" method of the item*:: a journal entry *where each journal entry is an HTMLJournalEntry.* Defining new web actions ------------------------ You may define new actions to be triggered by the ``@action`` form variable. These are added to the tracker ``interfaces.py`` as methods on the ``Client`` class. Adding action methods takes three steps; first you `define the new action method`_, then you `register the action method`_ with the cgi interface so it may be triggered by the ``@action`` form variable. Finally you `use the new action`_ in your HTML form. See "`setting up a "wizard" (or "druid") for controlled adding of issues`_" for an example. Define the new action method ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The action methods have the following interface:: def myActionMethod(self): ''' Perform some action. No return value is required. ''' The *self* argument is an instance of your tracker ``instance.Client`` class - thus it's mostly implemented by ``roundup.cgi.Client``. See the docstring of that class for details of what it can do. The method will typically check the ``self.form`` variable's contents. It may then: - add information to ``self.ok_message`` or ``self.error_message`` - change the ``self.template`` variable to alter what the user will see next - raise Unauthorised, SendStaticFile, SendFile, NotFound or Redirect exceptions Register the action method ~~~~~~~~~~~~~~~~~~~~~~~~~~ The method is now written, but isn't available to the user until you add it to the `instance.Client`` class ``actions`` variable, like so:: actions = client.Class.actions + ( ('myaction', 'myActionMethod'), ) This maps the action name "myaction" to the action method we defined. Use the new action ~~~~~~~~~~~~~~~~~~ In your HTML form, add a hidden form element like so:: where "myaction" is the name you registered in the previous step. Examples ======== .. contents:: :local: :depth: 1 Adding a new field to the classic schema ---------------------------------------- This example shows how to add a new constrained property (i.e. a selection of distinct values) to your tracker. Introduction ~~~~~~~~~~~~ To make the classic schema of roundup useful as a TODO tracking system for a group of systems administrators, it needed an extra data field per issue: a category. This would let sysadmins quickly list all TODOs in their particular area of interest without having to do complex queries, and without relying on the spelling capabilities of other sysadmins (a losing proposition at best). Adding a field to the database ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This is the easiest part of the change. The category would just be a plain string, nothing fancy. To change what is in the database you need to add some lines to the ``open()`` function in ``dbinit.py``. Under the comment:: # add any additional database schema configuration here add:: category = Class(db, "category", name=String()) category.setkey("name") Here we are setting up a chunk of the database which we are calling "category". It contains a string, which we are refering to as "name" for lack of a more imaginative title. (Since "name" is one of the properties that Roundup looks for on items if you do not set a key for them, it's probably a good idea to stick with it for new classes if at all appropriate.) Then we are setting the key of this chunk of the database to be that "name". This is equivalent to an index for database types. This also means that there can only be one category with a given name. Adding the above lines allows us to create categories, but they're not tied to the issues that we are going to be creating. It's just a list of categories off on its own, which isn't much use. We need to link it in with the issues. To do that, find the lines in the ``open()`` function in ``dbinit.py`` which set up the "issue" class, and then add a link to the category:: issue = IssueClass(db, "issue", ... , category=Multilink("category"), ... ) The ``Multilink()`` means that each issue can have many categories. If you were adding something with a one-to-one relationship to issues (such as the "assignedto" property), use ``Link()`` instead. That is all you need to do to change the schema. The rest of the effort is fiddling around so you can actually use the new category. Populating the new category class ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ If you haven't initialised the database with the roundup-admin "initialise" command, then you can add the following to the tracker ``dbinit.py`` in the ``init()`` function under the comment:: # add any additional database create steps here - but only if you # haven't initialised the database with the admin "initialise" command Add:: category = db.getclass('category') category.create(name="scipy", order="1") category.create(name="chaco", order="2") category.create(name="weave", order="3") If the database has already been initalised, then you need to use the ``roundup-admin`` tool:: % roundup-admin -i Roundup ready for input. Type "help" for help. roundup> create category name=scipy order=1 1 roundup> create category name=chaco order=1 2 roundup> create category name=weave order=1 3 roundup> exit... There are unsaved changes. Commit them (y/N)? y TODO: explain why order=1 in each case. Also, does key get set to "name" automatically when added via roundup-admin? Setting up security on the new objects ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ By default only the admin user can look at and change objects. This doesn't suit us, as we want any user to be able to create new categories as required, and obviously everyone needs to be able to view the categories of issues for it to be useful. We therefore need to change the security of the category objects. This is also done in the ``open()`` function of ``dbinit.py``. There are currently two loops which set up permissions and then assign them to various roles. Simply add the new "category" to both lists:: # new permissions for this schema for cl in 'issue', 'file', 'msg', 'user', 'category': db.security.addPermission(name="Edit", klass=cl, description="User is allowed to edit "+cl) db.security.addPermission(name="View", klass=cl, description="User is allowed to access "+cl) # Assign the access and edit permissions for issue, file and message # to regular users now for cl in 'issue', 'file', 'msg', 'category': p = db.security.getPermission('View', cl) db.security.addPermissionToRole('User', p) p = db.security.getPermission('Edit', cl) db.security.addPermissionToRole('User', p) So you are in effect doing the following (with 'cl' substituted by its value):: db.security.addPermission(name="Edit", klass='category', description="User is allowed to edit "+'category') db.security.addPermission(name="View", klass='category', description="User is allowed to access "+'category') which is creating two permission types; that of editing and viewing "category" objects respectively. Then the following lines assign those new permissions to the "User" role, so that normal users can view and edit "category" objects:: p = db.security.getPermission('View', 'category') db.security.addPermissionToRole('User', p) p = db.security.getPermission('Edit', 'category') db.security.addPermissionToRole('User', p) This is all the work that needs to be done for the database. It will store categories, and let users view and edit them. Now on to the interface stuff. Changing the web left hand frame ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We need to give the users the ability to create new categories, and the place to put the link to this functionality is in the left hand function bar, under the "Issues" area. The file that defines how this area looks is ``html/page``, which is what we are going to be editing next. If you look at this file you can see that it contains a lot of "classblock" sections which are chunks of HTML that will be included or excluded in the output depending on whether the condition in the classblock is met. Under the end of the classblock for issue is where we are going to add the category code::

Categories
New Category

The first two lines is the classblock definition, which sets up a condition that only users who have "View" permission for the "category" object will have this section included in their output. Next comes a plain "Categories" header in bold. Everyone who can view categories will get that. Next comes the link to the editing area of categories. This link will only appear if the condition - that the user has "Edit" permissions for the "category" objects - is matched. If they do have permission then they will get a link to another page which will let the user add new categories. Note that if you have permission to *view* but not to *edit* categories, then all you will see is a "Categories" header with nothing underneath it. This is obviously not very good interface design, but will do for now. I just claim that it is so I can add more links in this section later on. However to fix the problem you could change the condition in the classblock statement, so that only users with "Edit" permission would see the "Categories" stuff. Setting up a page to edit categories ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We defined code in the previous section which let users with the appropriate permissions see a link to a page which would let them edit conditions. Now we have to write that page. The link was for the *item* template of the *category* object. This translates into Roundup looking for a file called ``category.item.html`` in the ``html`` tracker directory. This is the file that we are going to write now. First we add an info tag in a comment which doesn't affect the outcome of the code at all, but is useful for debugging. If you load a page in a browser and look at the page source, you can see which sections come from which files by looking for these comments:: Next we need to add in the METAL macro stuff so we get the normal page trappings:: Category editing

Category editing

Next we need to setup up a standard HTML form, which is the whole purpose of this file. We link to some handy javascript which sends the form through only once. This is to stop users hitting the send button multiple times when they are impatient and thus having the form sent multiple times::
Next we define some code which sets up the minimum list of fields that we require the user to enter. There will be only one field - "name" - so they better put something in it, otherwise the whole form is pointless:: To get everything to line up properly we will put everything in a table, and put a nice big header on it so the user has an idea what is happening:: Next, we need the field into which the user is going to enter the new category. The "context.name.field(size=60)" bit tells Roundup to generate a normal HTML field of size 60, and the contents of that field will be the "name" variable of the current context (which is "category"). The upshot of this is that when the user types something in to the form, a new category will be created with that name:: Then a submit button so that the user can submit the new category:: Finally we finish off the tags we used at the start to do the METAL stuff:: So putting it all together, and closing the table and form we get:: Category editing This is quite a lot to just ask the user one simple question, but there is a lot of setup for basically one line (the form line) to do its work. To add another field to "category" would involve one more line (well, maybe a few extra to get the formatting correct). Adding the category to the issue ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We now have the ability to create issues to our heart's content, but that is pointless unless we can assign categories to issues. Just like the ``html/category.item.html`` file was used to define how to add a new category, the ``html/issue.item.html`` is used to define how a new issue is created. Just like ``category.issue.html`` this file defines a form which has a table to lay things out. It doesn't matter where in the table we add new stuff, it is entirely up to your sense of aesthetics:: First, we define a nice header so that the user knows what the next section is, then the middle line does what we are most interested in. This ``context/category/field`` gets replaced by a field which contains the category in the current context (the current context being the new issue). The classhelp lines generate a link (labelled "list") to a popup window which contains the list of currently known categories. Searching on categories ~~~~~~~~~~~~~~~~~~~~~~~ We can add categories, and create issues with categories. The next obvious thing that we would like to be able to do, would be to search for issues based on their category, so that, for example, anyone working on the web server could look at all issues in the category "Web". If you look for "Search Issues" in the 'html/page.html' file, you will find that it looks something like ``Search Issues``. This shows us that when you click on "Search Issues" it will be looking for a ``issue.search.html`` file to display. So that is the file that we will change. If you look at this file it should be starting to seem familiar, although it does use some new macros. You can add the new category search code anywhere you like within that form:: The definitions in the opening tag are used by the macros: - search_select expands to a drop-down box with all categories using db_klass and db_content. - column_input expands to a checkbox for selecting what columns should be displayed. - sort_input expands to a radio button for selecting what property should be sorted on. - group_input expands to a radio button for selecting what property should be group on. The category search code above would expand to the following:: Adding category to the default view ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We can now add categories, add issues with categories, and search for issues based on categories. This is everything that we need to do; however, there is some more icing that we would like. I think the category of an issue is important enough that it should be displayed by default when listing all the issues. Unfortunately, this is a bit less obvious than the previous steps. The code defining how the issues look is in ``html/issue.index.html``. This is a large table with a form down at the bottom for redisplaying and so forth. Firstly we need to add an appropriate header to the start of the table:: The *condition* part of this statement is to avoid displaying the Category column if the user has selected not to see it. The rest of the table is a loop which will go through every issue that matches the display criteria. The loop variable is "i" - which means that every issue gets assigned to "i" in turn. The new part of code to display the category will look like this:: The condition is the same as above: only display the condition when the user hasn't asked for it to be hidden. The next part is to set the content of the cell to be the category part of "i" - the current issue. Finally we have to edit ``html/page.html`` again. This time, we need to tell it that when the user clicks on "Unasigned Issues" or "All Issues", the category column should be included in the resulting list. If you scroll down the page file, you can see the links with lots of options. The option that we are interested in is the ``:columns=`` one which tells roundup which fields of the issue to display. Simply add "category" to that list and it all should work. Adding in state transition control ---------------------------------- Sometimes tracker admins want to control the states that users may move issues to. You can do this by following these steps: 1. make "status" a required variable. This is achieved by adding the following to the top of the form in the ``issue.item.html`` template:: this will force users to select a status. 2. add a Multilink property to the status class:: stat = Class(db, "status", ... , transitions=Multilink('status'), ...) and then edit the statuses already created, either: a. through the web using the class list -> status class editor, or b. using the roundup-admin "set" command. 3. add an auditor module ``checktransition.py`` in your tracker's ``detectors`` directory, for example:: def checktransition(db, cl, nodeid, newvalues): ''' Check that the desired transition is valid for the "status" property. ''' if not newvalues.has_key('status'): return current = cl.get(nodeid, 'status') new = newvalues['status'] if new == current: return ok = db.status.get(current, 'transitions') if new not in ok: raise ValueError, 'Status not allowed to move from "%s" to "%s"'%( db.status.get(current, 'name'), db.status.get(new, 'name')) def init(db): db.issue.audit('set', checktransition) 4. in the ``issue.item.html`` template, change the status editing bit from:: to:: which displays only the allowed status to transition to. Displaying only message summaries in the issue display ------------------------------------------------------ Alter the issue.item template section for messages to::
Category
Name name
  submit button will go here

Category editing
Category
Name name
  submit button will go here
Category
Priority:
Category:
CategoryStatus statusStatus
Messages
author date summary remove
Restricting the list of users that are assignable to a task ----------------------------------------------------------- 1. In your tracker's "dbinit.py", create a new Role, say "Developer":: db.security.addRole(name='Developer', description='A developer') 2. Just after that, create a new Permission, say "Fixer", specific to "issue":: p = db.security.addPermission(name='Fixer', klass='issue', description='User is allowed to be assigned to fix issues') 3. Then assign the new Permission to your "Developer" Role:: db.security.addPermissionToRole('Developer', p) 4. In the issue item edit page ("html/issue.item.html" in your tracker directory), use the new Permission in restricting the "assignedto" list:: For extra security, you may wish to setup an auditor to enforce the Permission requirement (install this as "assignedtoFixer.py" in your tracker "detectors" directory):: def assignedtoMustBeFixer(db, cl, nodeid, newvalues): ''' Ensure the assignedto value in newvalues is a used with the Fixer Permission ''' if not newvalues.has_key('assignedto'): # don't care return # get the userid userid = newvalues['assignedto'] if not db.security.hasPermission('Fixer', userid, cl.classname): raise ValueError, 'You do not have permission to edit %s'%cl.classname def init(db): db.issue.audit('set', assignedtoMustBeFixer) db.issue.audit('create', assignedtoMustBeFixer) So now, if an edit action attempts to set "assignedto" to a user that doesn't have the "Fixer" Permission, the error will be raised. Setting up a "wizard" (or "druid") for controlled adding of issues ------------------------------------------------------------------ 1. Set up the page templates you wish to use for data input. My wizard is going to be a two-step process: first figuring out what category of issue the user is submitting, and then getting details specific to that category. The first page includes a table of help, explaining what the category names mean, and then the core of the form::
Category: The next page has the usual issue entry information, with the addition of the following form fragments::
. . .
Note that later in the form, I test the value of "cat" include form elements that are appropriate. For example:: Operating System Web Browser ... the above section will only be displayed if the category is one of 6, 10, 13, 14, 15, 16 or 17. 3. Determine what actions need to be taken between the pages - these are usually to validate user choices and determine what page is next. Now encode those actions in methods on the ``interfaces.Client`` class and insert hooks to those actions in the "actions" attribute on that class, like so:: actions = client.Client.actions + ( ('page1_submit', 'page1SubmitAction'), ) def page1SubmitAction(self): ''' Verify that the user has selected a category, and then move on to page 2. ''' category = self.form['category'].value if category == '-1': self.error_message.append('You must select a category of report') return # everything's ok, move on to the next page self.template = 'add_page2' 4. Use the usual "new" action as the ``@action`` on the final page, and you're done (the standard context/submit method can do this for you). Using an external password validation source -------------------------------------------- We have a centrally-managed password changing system for our users. This results in a UN*X passwd-style file that we use for verification of users. Entries in the file consist of ``name:password`` where the password is encrypted using the standard UN*X ``crypt()`` function (see the ``crypt`` module in your Python distribution). An example entry would be:: admin:aamrgyQfDFSHw Each user of Roundup must still have their information stored in the Roundup database - we just use the passwd file to check their password. To do this, we add the following code to our ``Client`` class in the tracker home ``interfaces.py`` module:: def verifyPassword(self, userid, password): # get the user's username username = self.db.user.get(userid, 'username') # the passwords are stored in the "passwd.txt" file in the # tracker home file = os.path.join(self.db.config.TRACKER_HOME, 'passwd.txt') # see if we can find a match for ent in [line.strip().split(':') for line in open(file).readlines()]: if ent[0] == username: return crypt.crypt(password, ent[1][:2]) == ent[1] # user doesn't exist in the file return 0 What this does is look through the file, line by line, looking for a name that matches. We also remove the redundant password fields from the ``user.item`` template. Adding a "vacation" flag to users for stopping nosy messages ------------------------------------------------------------ When users go on vacation and set up vacation email bouncing, you'll start to see a lot of messages come back through Roundup "Fred is on vacation". Not very useful, and relatively easy to stop. 1. add a "vacation" flag to your users:: user = Class(db, "user", username=String(), password=Password(), address=String(), realname=String(), phone=String(), organisation=String(), alternate_addresses=String(), roles=String(), queries=Multilink("query"), vacation=Boolean()) 2. So that users may edit the vacation flags, add something like the following to your ``user.item`` template:: On Vacation vacation 3. edit your detector ``nosyreactor.py`` so that the ``nosyreaction()`` consists of:: def nosyreaction(db, cl, nodeid, oldvalues): # send a copy of all new messages to the nosy list for msgid in determineNewMessages(cl, nodeid, oldvalues): try: users = db.user messages = db.msg # figure the recipient ids sendto = [] r = {} recipients = messages.get(msgid, 'recipients') for recipid in messages.get(msgid, 'recipients'): r[recipid] = 1 # figure the author's id, and indicate they've received # the message authid = messages.get(msgid, 'author') # possibly send the message to the author, as long as # they aren't anonymous if (db.config.MESSAGES_TO_AUTHOR == 'yes' and users.get(authid, 'username') != 'anonymous'): sendto.append(authid) r[authid] = 1 # now figure the nosy people who weren't recipients nosy = cl.get(nodeid, 'nosy') for nosyid in nosy: # Don't send nosy mail to the anonymous user (that # user shouldn't appear in the nosy list, but just # in case they do...) if users.get(nosyid, 'username') == 'anonymous': continue # make sure they haven't seen the message already if not r.has_key(nosyid): # send it to them sendto.append(nosyid) recipients.append(nosyid) # generate a change note if oldvalues: note = cl.generateChangeNote(nodeid, oldvalues) else: note = cl.generateCreateNote(nodeid) # we have new recipients if sendto: # filter out the people on vacation sendto = [i for i in sendto if not users.get(i, 'vacation', 0)] # map userids to addresses sendto = [users.get(i, 'address') for i in sendto] # update the message's recipients list messages.set(msgid, recipients=recipients) # send the message cl.send_message(nodeid, msgid, note, sendto) except roundupdb.MessageSendError, message: raise roundupdb.DetectorError, message Note that this is the standard nosy reaction code, with the small addition of:: # filter out the people on vacation sendto = [i for i in sendto if not users.get(i, 'vacation', 0)] which filters out the users that have the vacation flag set to true. Adding a time log to your issues -------------------------------- We want to log the dates and amount of time spent working on issues, and be able to give a summary of the total time spent on a particular issue. 1. Add a new class to your tracker ``dbinit.py``:: # storage for time logging timelog = Class(db, "timelog", period=Interval()) Note that we automatically get the date of the time log entry creation through the standard property "creation". 2. Link to the new class from your issue class (again, in ``dbinit.py``):: issue = IssueClass(db, "issue", assignedto=Link("user"), topic=Multilink("keyword"), priority=Link("priority"), status=Link("status"), times=Multilink("timelog")) the "times" property is the new link to the "timelog" class. 3. We'll need to let people add in times to the issue, so in the web interface we'll have a new entry field. This is a special field because unlike the other fields in the issue.item template, it affects a different item (a timelog item) and not the template's item, an issue. We have a special syntax for form fields that affect items other than the template default item (see the cgi documentation on `special form variables`_). In particular, we add a field to capture a new timelog item's perdiod:: Time Log
(enter as '3y 1m 4d 2:40:02' or parts thereof) and another hidden field that links that new timelog item (new because it's marked as having id "-1") to the issue item. It looks like this:: On submission, the "-1" timelog item will be created and assigned a real item id. The "times" property of the issue will have the new id added to it. 4. We want to display a total of the time log times that have been accumulated for an issue. To do this, we'll need to actually write some Python code, since it's beyond the scope of PageTemplates to perform such calculations. We do this by adding a method to the TemplatingUtils class in our tracker ``interfaces.py`` module:: class TemplatingUtils: ''' Methods implemented on this class will be available to HTML templates through the 'utils' variable. ''' def totalTimeSpent(self, times): ''' Call me with a list of timelog items (which have an Interval "period" property) ''' total = Interval('') for time in times: total += time.period._value return total Replace the ``pass`` line as we did in step 4 above with the Client class. As indicated in the docstrings, we will be able to access the ``totalTimeSpent`` method via the ``utils`` variable in our templates. 5. Display the time log for an issue::
Time Log
DatePeriodLogged By
I put this just above the Messages log in my issue display. Note our use of the ``totalTimeSpent`` method which will total up the times for the issue and return a new Interval. That will be automatically displayed in the template as text like "+ 1y 2:40" (1 year, 2 hours and 40 minutes). 8. If you're using a persistent web server - roundup-server or mod_python for example - then you'll need to restart that to pick up the code changes. When that's done, you'll be able to use the new time logging interface. Using a UN*X passwd file as the user database --------------------------------------------- On some systems the primary store of users is the UN*X passwd file. It holds information on users such as their username, real name, password and primary user group. Roundup can use this store as its primary source of user information, but it needs additional information too - email address(es), roundup Roles, vacation flags, roundup hyperdb item ids, etc. Also, "retired" users must still exist in the user database, unlike some passwd files in which the users are removed when they no longer have access to a system. To make use of the passwd file, we therefore synchronise between the two user stores. We also use the passwd file to validate the user logins, as described in the previous example, `using an external password validation source`_. We keep the users lists in sync using a fairly simple script that runs once a day, or several times an hour if more immediate access is needed. In short, it: 1. parses the passwd file, finding usernames, passwords and real names, 2. compares that list to the current roundup user list: a. entries no longer in the passwd file are *retired* b. entries with mismatching real names are *updated* c. entries only exist in the passwd file are *created* 3. send an email to administrators to let them know what's been done. The retiring and updating are simple operations, requiring only a call to ``retire()`` or ``set()``. The creation operation requires more information though - the user's email address and their roundup Roles. We're going to assume that the user's email address is the same as their login name, so we just append the domain name to that. The Roles are determined using the passwd group identifier - mapping their UN*X group to an appropriate set of Roles. The script to perform all this, broken up into its main components, is as follows. Firstly, we import the necessary modules and open the tracker we're to work on:: import sys, os, smtplib from roundup import instance, date # open the tracker tracker_home = sys.argv[1] tracker = instance.open(tracker_home) Next we read in the *passwd* file from the tracker home:: # read in the users file = os.path.join(tracker_home, 'users.passwd') users = [x.strip().split(':') for x in open(file).readlines()] Handle special users (those to ignore in the file, and those who don't appear in the file):: # users to not keep ever, pre-load with the users I know aren't # "real" users ignore = ['ekmmon', 'bfast', 'csrmail'] # users to keep - pre-load with the roundup-specific users keep = ['comment_pool', 'network_pool', 'admin', 'dev-team', 'cs_pool', 'anonymous', 'system_pool', 'automated'] Now we map the UN*X group numbers to the Roles that users should have:: roles = { '501': 'User,Tech', # tech '502': 'User', # finance '503': 'User,CSR', # customer service reps '504': 'User', # sales '505': 'User', # marketing } Now we do all the work. Note that the body of the script (where we have the tracker database open) is wrapped in a ``try`` / ``finally`` clause, so that we always close the database cleanly when we're finished. So, we now do all the work:: # open the database db = tracker.open('admin') try: # store away messages to send to the tracker admins msg = [] # loop over the users list read in from the passwd file for user,passw,uid,gid,real,home,shell in users: if user in ignore: # this user shouldn't appear in our tracker continue keep.append(user) try: # see if the user exists in the tracker uid = db.user.lookup(user) # yes, they do - now check the real name for correctness if real != db.user.get(uid, 'realname'): db.user.set(uid, realname=real) msg.append('FIX %s - %s'%(user, real)) except KeyError: # nope, the user doesn't exist db.user.create(username=user, realname=real, address='%s@ekit-inc.com'%user, roles=roles[gid]) msg.append('ADD %s - %s (%s)'%(user, real, roles[gid])) # now check that all the users in the tracker are also in our # "keep" list - retire those who aren't for uid in db.user.list(): user = db.user.get(uid, 'username') if user not in keep: db.user.retire(uid) msg.append('RET %s'%user) # if we did work, then send email to the tracker admins if msg: # create the email msg = '''Subject: %s user database maintenance %s '''%(db.config.TRACKER_NAME, '\n'.join(msg)) # send the email smtp = smtplib.SMTP(db.config.MAILHOST) addr = db.config.ADMIN_EMAIL smtp.sendmail(addr, addr, msg) # now we're done - commit the changes db.commit() finally: # always close the database cleanly db.close() And that's it! Using an LDAP database for user information ------------------------------------------- A script that reads users from an LDAP store using http://python-ldap.sf.net/ and then compares the list to the users in the roundup user database would be pretty easy to write. You'd then have it run once an hour / day (or on demand if you can work that into your LDAP store workflow). See the example `Using a UN*X passwd file as the user database`_ for more information about doing this. To authenticate off the LDAP store (rather than using the passwords in the roundup user database) you'd use the same python-ldap module inside an extension to the cgi interface. You'd do this by adding a method called "verifyPassword" to the Client class in your tracker's interfaces.py module. The method is implemented by default as:: def verifyPassword(self, userid, password): ''' Verify the password that the user has supplied ''' stored = self.db.user.get(self.userid, 'password') if password == stored: return 1 if not password and not stored: return 1 return 0 So you could reimplement this as something like:: def verifyPassword(self, userid, password): ''' Verify the password that the user has supplied ''' # look up some unique LDAP information about the user username = self.db.user.get(self.userid, 'username') # now verify the password supplied against the LDAP store Enabling display of either message summaries or the entire messages ------------------------------------------------------------------- This is pretty simple - all we need to do is copy the code from the example `displaying only message summaries in the issue display`_ into our template alongside the summary display, and then introduce a switch that shows either one or the other. We'll use a new form variable, ``@whole_messages`` to achieve this::
Messages show entire messages
author date summary remove
Messages show only summaries
author date (remove)
Blocking issues that depend on other issues ------------------------------------------- We needed the ability to mark certain issues as "blockers" - that is, they can't be resolved until another issue (the blocker) they rely on is resolved. To achieve this: 1. Create a new property on the issue Class, ``blockers=Multilink("issue")``. Edit your tracker's dbinit.py file. Where the "issue" class is defined, something like:: issue = IssueClass(db, "issue", assignedto=Link("user"), topic=Multilink("keyword"), priority=Link("priority"), status=Link("status")) add the blockers entry like so:: issue = IssueClass(db, "issue", blockers=Multilink("issue"), assignedto=Link("user"), topic=Multilink("keyword"), priority=Link("priority"), status=Link("status")) 2. Add the new "blockers" property to the issue.item edit page, using something like:: Waiting On
View: You'll need to fiddle with your item page layout to find an appropriate place to put it - I'll leave that fun part up to you. Just make sure it appears in the first table, possibly somewhere near the "superseders" field. 3. Create a new detector module (attached) which enforces the rules: - issues may not be resolved if they have blockers - when a blocker is resolved, it's removed from issues it blocks The contents of the detector should be something like this:: def blockresolution(db, cl, nodeid, newvalues): ''' If the issue has blockers, don't allow it to be resolved. ''' if nodeid is None: blockers = [] else: blockers = cl.get(nodeid, 'blockers') blockers = newvalues.get('blockers', blockers) # don't do anything if there's no blockers or the status hasn't # changed if not blockers or not newvalues.has_key('status'): return # get the resolved state ID resolved_id = db.status.lookup('resolved') # format the info u = db.config.TRACKER_WEB s = ', '.join(['%s'%( u,id,id) for id in blockers]) if len(blockers) == 1: s = 'issue %s is'%s else: s = 'issues %s are'%s # ok, see if we're trying to resolve if newvalues['status'] == resolved_id: raise ValueError, "This issue can't be resolved until %s resolved."%s def resolveblockers(db, cl, nodeid, newvalues): ''' When we resolve an issue that's a blocker, remove it from the blockers list of the issue(s) it blocks. ''' if not newvalues.has_key('status'): return # get the resolved state ID resolved_id = db.status.lookup('resolved') # interesting? if newvalues['status'] != resolved_id: return # yes - find all the blocked issues, if any, and remove me from # their blockers list issues = cl.find(blockers=nodeid) for issueid in issues: blockers = cl.get(issueid, 'blockers') if nodeid in blockers: blockers.remove(nodeid) cl.set(issueid, blockers=blockers) def init(db): # might, in an obscure situation, happen in a create db.issue.audit('create', blockresolution) db.issue.audit('set', blockresolution) # can only happen on a set db.issue.react('set', resolveblockers) Put the above code in a file called "blockers.py" in your tracker's "detectors" directory. 4. Finally, and this is an optional step, modify the tracker web page URLs so they filter out issues with any blockers. You do this by adding an additional filter on "blockers" for the value "-1". For example, the existing "Show All" link in the "page" template (in the tracker's "html" directory) looks like this:: Show All
modify it to add the "blockers" info to the URL (note, both the ":filter" *and* "blockers" values must be specified):: Show All
That's it. You should now be able to set blockers on your issues. Note that if you want to know whether an issue has any other issues dependent on it (i.e. it's in their blockers list) you can look at the journal history at the bottom of the issue page - look for a "link" event to another issue's "blockers" property. Add users to the nosy list based on the topic --------------------------------------------- We need the ability to automatically add users to the nosy list based on the occurence of a topic. Every user should be allowed to edit his own list of topics for which he wants to be added to the nosy list. Below will be showed that such a change can be performed with only minimal understanding of the roundup system, but with clever use of Copy and Paste. This requires three changes to the tracker: a change in the database to allow per-user recording of the lists of topics for which he wants to be put on the nosy list, a change in the user view allowing to edit this list of topics, and addition of an auditor which updates the nosy list when a topic is set. Adding the nosy topic list ~~~~~~~~~~~~~~~~~~~~~~~~~~ The change in the database to make is that for any user there should be a list of topics for which he wants to be put on the nosy list. Adding a ``Multilink`` of ``keyword`` seem to fullfill this (note that within the code topics are called ``keywords``.) As such, all what has to be done is to add a new field to the definition of ``user`` within the file ``dbinit.py``. We will call this new field ``nosy_keywords``, and the updated definition of user will be:: user = Class(db, "user", username=String(), password=Password(), address=String(), realname=String(), phone=String(), organisation=String(), alternate_addresses=String(), queries=Multilink('query'), roles=String(), timezone=String(), nosy_keywords=Multilink('keyword')) Changing the user view to allow changing the nosy topic list ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We want any user to be able to change the list of topics for which he will by default be added to the nosy list. We choose to add this to the user view, as is generated by the file ``html/user.item.html``. We easily can see that the topic field in the issue view has very similar editting requirements as our nosy topics, both being a list of topics. As such, we search for Topics in ``issue.item.html``, and extract the associated parts from there. We add this to ``user.item.html`` at the bottom of the list of viewed items (i.e. just below the 'Alternate E-mail addresses' in the classic template):: Nosy Topics Addition of an auditor to update the nosy list ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The more difficult part is the addition of the logic to actually at the users to the nosy list when it is required. The choice is made to perform this action when the topics on an item are set, including when an item is created. Here we choose to start out with a copy of the ``detectors/nosyreaction.py`` detector, which we copy to the file ``detectors/nosy_keyword_reaction.py``. This looks like a good start as it also adds users to the nosy list. A look through the code reveals that the ``nosyreaction`` function actually is sending the e-mail, which we do not need. As such, we can change the init function to:: def init(db): db.issue.audit('create', update_kw_nosy) db.issue.audit('set', update_kw_nosy) After that we rename the ``updatenosy`` function to ``update_kw_nosy``. The first two blocks of code in that function relate to settings ``current`` to a combination of the old and new nosy lists. This functionality is left in the new auditor. The following block of code, which in ``updatenosy`` handled adding the assignedto user(s) to the nosy list, should be replaced by a block of code to add the interested users to the nosy list. We choose here to loop over all new topics, than loop over all users, and assign the user to the nosy list when the topic in the user's nosy_keywords. The next part in ``updatenosy``, adding the author and/or recipients of a message to the nosy list, obviously is not relevant here and thus is deleted from the new auditor. The last part, copying the new nosy list to newvalues, does not have to be changed. This brings the following function:: def update_kw_nosy(db, cl, nodeid, newvalues): '''Update the nosy list for changes to the topics ''' # nodeid will be None if this is a new node current = {} if nodeid is None: ok = ('new', 'yes') else: ok = ('yes',) # old node, get the current values from the node if they haven't # changed if not newvalues.has_key('nosy'): nosy = cl.get(nodeid, 'nosy') for value in nosy: if not current.has_key(value): current[value] = 1 # if the nosy list changed in this transaction, init from the new value if newvalues.has_key('nosy'): nosy = newvalues.get('nosy', []) for value in nosy: if not db.hasnode('user', value): continue if not current.has_key(value): current[value] = 1 # add users with topic in nosy_keywords to the nosy list if newvalues.has_key('topic') and newvalues['topic'] is not None: topic_ids = newvalues['topic'] for topic in topic_ids: # loop over all users, # and assign user to nosy when topic in nosy_keywords for user_id in db.user.list(): nosy_kw = db.user.get(user_id, "nosy_keywords") found = 0 for kw in nosy_kw: if kw == topic: found = 1 if found: current[user_id] = 1 # that's it, save off the new nosy list newvalues['nosy'] = current.keys() and these two function are the only ones needed in the file. TODO: update this example to use the find() Class method. Caveats ~~~~~~~ A few problems with the design here can be noted: Multiple additions When a user, after automatic selection, is manually removed from the nosy list, he again is added to the nosy list when the topic list of the issue is updated. A better design might be to only check which topics are new compared to the old list of topics, and only add users when they have indicated interest on a new topic. The code could also be changed to only trigger on the create() event, rather than also on the set() event, thus only setting the nosy list when the issue is created. Scalability In the auditor there is a loop over all users. For a site with only few users this will pose no serious problem, however, with many users this will be a serious performance bottleneck. A way out will be to link from the topics to the users which selected these topics a nosy topics. This will eliminate the loop over all users. Adding action links to the index page ------------------------------------- Add a column to the item.index.html template. Resolving the issue:: resolve "Take" the issue:: take ... and so on ------------------- Back to `Table of Contents`_ .. _`Table of Contents`: index.html .. _`design documentation`: design.html