Summary of OpenFormula JIRA meeting 2009-11-10

["David A. Wheeler" <dwheeler@dwheeler.com>]

Here's my summary of the 2009-11-10 OpenFormula comment processing (JIRA) teleconference.

Participants:
Rob Weir
Eric Patterson
Andreas
David A. Wheeler
Dennis Hamilton
[Clarification: Eike was *not* on this call.]

The agenda, as usual, was:
* Controversial issues & proposed resolutions needing discussion
* Allocation of unassigned JIRA comments

First, we discussed controversial issues that needed group discussion.

There was a brief discussion on exponentiation (e.g., POWER vs. ^).
They can't be made exactly the same; this has a potential ripple effect on complex numbers.

Weir: We just need to finish what we need to finish up,
then post it into the TC repository.  I know Patrick D. looked it over,
but once part 1 is done, can he focus on this draft specification.

There was a brief discussion on the test cases.
The final spec version will NOT include the test cases, but they
could be included in other formats.
Rob noted that in cross-implementation testing
of the "Small" group, implementations did very well in "normal" cases;
the problems were the edge cases:
* Some implementations were failing to implement the maximum length of formula stated.
* Re: Count of error cells - different implementations do different things

Then there was a long discussion - should we change how we handle maximum length of a formula?
Wheeler: Before we change that, let's find out who does not meet it, & how difficult it would be to do.
Weir: As the maximum length gets longer, it starts to basically be abuse of spreadsheets.
Note that Excel 2007 increased the maximum formula length to 16K.
It's actually more complicated in Excel... internally some expressions get
expanded internally (e.g., names).
If the formula is too large for the implementation, what should the spec say?
The implementation shouldn't truncate silently.

Eric: 1024 is actually not THAT long;
you can get close to the limit without too much trouble.
We found that changing from 1Ki to 16Ki was a LOT of work
for the Excel implementation.
When saving to another format (where they're lost), Excel gives a warning.

Weir/Eric: Change the OpenFormula specification of length.
Remove the "[..]" processing, and make it a simple maximum length.
The group agreed to change the maximum to a simple 2048 characters.
(Implementations must support at LEAST this size.)

We then allocated unassigned issues:
846: Dennis
1191: Dennis
1194: Rob Weir (tied to 1196)
1076: Rob Weir
1911: Dennis
2014: Weir (really a final pass review)
2073: Dennis.  (Hopefully this is resolved now, need to double-check).

1999 (this adds a new function that generates text for numbers in arbitrary languages) will be reassigned to ODF-NEXT.  This looks useful, but isn't widely implemented.

Next call: Need to discuss Calculation settings & regular expression syntax.

Next call will be in 2 weeks (as usual), 10am U.S. Eastern Time.
Wheeler *believes* that by then all Northern Hemisphere countries
that use daylight savings time will be re-synchronized.

The next meeting will be near Thanksgiving (U.S. holiday);
Rob Weir will be on vacation, but others won't, so we'll carry on.

--- David A. Wheeler