From Glom
< Development
Revision as of 15:42, 4 December 2007 by Armin (Talk | contribs)

Jump to: navigation, search

Please consider funding my work on Glom so that I can implement all these features. Murray Cumming

Or you could just donate via PayPal: x-click-but04.gif

Many thanks to our Supporters so far.

See also:

Post-1.6 Features

Easier tasks

  • For instance, creating the initial user, after asking for the root password.
  • This can be a completely separate program. It can be in any programming language (Python with pygtk might be appropriate).
  • It should parse the postgres configuration files, and then change them where necessary.
  • To ask for, and gain, root (or superuser) rights, do whatever synaptic or gnome-system-tools does. There is commented-out code to get root rights in this way already, in
  • Warn when deleting used fields and relationships: If a field or relationship is on a layout or report, or a field is in a relationship, warn before deleting.
  • Create a new warning dialog in the Glade file, and create a C++ class for it.
  • Add methods to the Document_Glom class, such as bool get_field_used(). Extra output parameters could return a string describing the use, or specific information about the layout, relationship, report, etc, on which it is used.
  • Whenever the UI allows someone to delete one of these things, call the method and show a dialog if necessary.
  • Depending on the dialog response, continue or cancel the deletion.
  • Quickly add Relationship/Field: For instance, when adding a field to a layout, quickly add a relationship so that you can add a field from it, without closing the layout window, opening the relationships window, and opening the layout window again. Or to quickly add a field to the current table when adding a field to the layout dialog. Or, when specifying a related choices lists (in a Field'a default formatting). (30 hours)
  • Add buttons to the dialogs in Glade.
  • Create signal handlers for those buttons in the dialog classes.
  • In the button signal handlers, open the appropriate dialogs - see the code that does the same from the regular menu signal handlers. Then update the current dialog to show the new available choices.
  • Ideally, think of a way to avoid this causing so many windows to be open at once, on top of each other.
  • Related Records as Calendar: Display a portal of related records in a calendar, using a specified field as the date.
  • Drag-and-drop layout: Add a panel of available items and show visual feedback as they are moved around on the layout. But still use the automatic-layout system.
  • Steal ideas from Glade. Time: 60 hours.
  • Windows Port: Shouldn't be too difficult. All dependencies are portable. Time: 60 hours.
  • This should use native GTK+ on Windows rather than X11 on Windows.
  • This should use g++, not MSCVC++.

Tasks requiring more experience

Some of these require you to have experience of, or learn about, the technologies marked in bold, such as Cairo, Avahi, or libgda.

  • Detect that an instance is open already, and connect to that computer, instead of re-hosting the database. Avahi can probably do this.
  • Detect that other clients are connected, and warn before closing.
  • Custom Print Layout: For instance, a perfect-looking official invoice. Implement with Cairo and the GTK+ 2.10 printing API. Time: 80 hours.
  • Automatic Server Detection: Patch postgres to use Avahi for ZeroConf/Bonjour/Rendezvous service broadcast. Postgres already uses Bonjour on MacOS X.
  • Deal with lots of data The custom tree model should progressively gets only the record values that are visible in the scrolled window. libgda 3.0, which we now use, has API that apparently allows this by supporting database cursors, but that is not actually implemented in the libgda postgres provider yet. (DONE)
  • Scripting: Most common actions should be standard features without the awkwardness of scripting, but maybe we need some way via the custom buttons to, for instance, on Contacts details create a related Invoice and take you to the Invoice details. For this, we must just present a python API for some parts of the Glom structure in addition to the current record object. This requires knowledge of the Python C API, which is used already for calculated fields. Alternatively, you may use boost::python if you are already familiar with it.
  • Non-Indexed Calculated Fields: Optionally don't store values - just recalc them when seen/used, and therefore don't allow their use in relationships, though that's awkwardly technical when seen in FileMaker. Time-to-complete: 60 hours.
  • Locking: Don't let a second user edit the same record. (Delay unlock for 5 seconds after entry loses focus). Update the display when the second user changes the same record. Time: 50 hours.
  • Multi-column relationships: Match on 2 key fields instead of just one. Time: 30 hours.
  • Performance:: Keep connection alive, do less by-value copying, more caching and hashing. libgda 2.0 (when finished) should help with this.
  • Web UI: Possible in principle, although it probably wouldn't be as responsive as a desktop application. All the needed information is in the .glom document. Something for a web developer, and there's no need to choose a favourite language/platform - just do it.
  • Embedded Glom: Port to Maemo and add new features to help.
  • Alien Tables: Allows some tables to get/set data in non-postgres sources, such as evolution-data-server, so people don't need to enter their contact information twice.
  • Multimedia Fields:
    • We currently have Image fields that can store images in almost any format, but the image is actually converted to PNG before saving in the database, so we always know what format the data has. To preserve original formats, and to support audio and video, we would need to store the mime type as well. We would need to define a binary format for the data, so we could store the mime type along with the data. An existing format would be best.
    • Then we would need to implement video and audio playing widgets, and we should offer to open the appropriate external application where we can't "play" the data inline.
    • Time: 80 hours.