Calc

One week of successful hacking at the Cambridge collaboration hackfest.

After I already struggled with Telepathy for some while, trying to implement DBus tubes over XMPP, Michael Meeks arranged a collaborative hackfest at Cambridge, where Collabora Ltd. is situated who to a large part are the maintainers of the Telepathy library. Nice to meet some of these guys (btw, if you happen to be in Cambridge and are into good Indian food go for the Golden Curry).

So last week Will Thompson from Collabora was set aside to work with us, who was of great help and immediately spotted why my tube acceptor never caught any offered tube, silly me, not spinning GMainLoop yet another time to wait for a specific state change ... Before, a connection to DBus could already be setup and a session between two contacts initiated and a tube offered, but I didn't get to the point where a tube was really established.

What we achieved from Tuesday to Thursday / Friday morning:

• use Empathy's XMPP account and contact list
• establish a DBus tube between an account and a contact
• transfer a file over the tube (which shall be the document to be worked on)
• intercept Calc input (basic proof of concept)
• transfer input as packet through tube
• receive on other end and use as input

Kudos again to Will, who made the fast progress on the Telepathy bits possible. And Michael for tearing apart Calc's ScDocFunc. That I now have to glue together again somehow ;-)

All happens on the LibreOffice feature/tubes2 branch. You're more than welcome to help out, especially if you're familiar with Telepathy or like to implement some dialogs in VCL such as an account/contact picker or have ideas how to improve our bits. You might also be interested that further work on this topic is offered as a Google Summer of Code (GSoC) project. There's plenty to do.

Btw, the next opportunity to collaboratively hack on collaboration would be at the Hamburg Hackfest on April 14/15 :-)

Update 2012-03-27T13:25+0200
Michael has some more details with bells and whistles and video over at his blog, video here:

Meet the LibreOffice team around the LibreOffice Devroom at FOSDEM 2012 room H.1309

Join my Calc Hacking for Starters session on Saturday 16:45, or how to not get trapped in or squashed between cells.

I've seen many (ab)uses of spreadsheets during my years as an OOo/LibO Calc developer, but this Spreadsheet as Music Tracker-Sequencer is the most hackish one.

You might also be interested in the How-To video.

[via Bjoern's My productivity suite sounds better than yours ...]

Abstract: Calc's (and in Writer table) cell input now needs to match locale dependent date acceptance patterns before it is recognized as a valid date.

Previously the number formatter's input scanner was very lax in what it accepted as a "valid" date. All combinations of 2-3 numbers separated by '.' '/' '-' or the locale's date separator even with blanks in between that somehow could be interpreted as a date was accepted as such, which was especially confusing with incomplete dates containing only 2 numbers that in many cases were meant as textual input instead. For example

• In en-US locale, M/D is a valid date input to be interpreted as day of month of current year. However, M/D/ and M.D. were accepted as well.
• In de-DE locale, D.M. is a valid date input to be interpreted as day of month of current year. However, D.M and D/M and D/M/ were accepted as well.

In case of an input like 1.2 in a de-DE locale or others using '.' separator, meant as some sort of textual numbering, this was extremely annoying, it was interpreted as 1st of February of current year and the user had to prepend a single quote / apostrophe to suppress date recognition. Similar for 1.2.3 in locales that do not use the '.' date separator.

Now, during build time for each locale one full date acceptance pattern is generated from the existing locale data's number format FormatElement with formatindex="21" that is also used to edit dates, taking the DMY order and the defined DateSeparator. For example, in the en-US locale this generates M/D/Y from the MM/DD/YYYY FormatCode, and in the de-DE locale D.M.Y from the DD.MM.YYYY code. For this to work correctly the separator used in the FormatCode must match the DateSeparator element defined in Separators. As for all rules there's one exception though ;) if the format code uses a different separator and that is one of the known '-' '.' '/' separators, a second pattern is generated using the format's separator. This as a generalized case for locales that for example may use an ISO 8601 edit format, as hu-HU does, regardless what the date separator is defined to.

Additionally to the date acceptance pattern every locale of course still accepts input in an ISO 8601 Y-M-D pattern, and since LibreOffice 3.5 that also leads to the YYYY-MM-DD format being applied.

Localizers, HEADS UP please

If in your locale incomplete dates should be accepted or additional patterns that vary from the generated full date pattern are needed, those are to be defined in the locale data LC_FORMAT element for which a new DateAcceptancePattern element exists, of which zero or more can occur before the FormatElement elements. Currently only the following patterns are defined as they are the only ones I knew were plausible:

• bg-BG, a trailing breaking or non-breaking space followed by lower case or upper case Cyrillic letter GHE and a dot, as defined in the edit format
  • D.M.Y г.
  • D.M.Y г.
  • D.M.Y Г.
  • D.M.Y Г.
• de-DE, incomplete date
  • D.M.
• en-US, incomplete date
  • M/D
• sl-SI, date separator dot plus space
  • D. M. Y

For example see i18npool/source/localedata/data/en_US.xml

Happy date accepting :-)

Update: an updated list of locales and patterns is available in a newer blog post.