Slashdot Mirror
Gnumeric Turns 5
Jody Goldberg writes "Five years ago, Miguel committed the first code for Gnumeric to CVS. In a testament to the quality of the code several lines are still in use. Since that time the project has grown to more than 300,000 lines and now supports all 325 worksheet functions in MS Excel, plus almost 100 more. This seemed like a good time to thank all the people who have contributed to Gnumeric over the years. We're about to start the run up to the the next stable release which will be out in a few
weeks and we look forward to continuing work with GNOME, and the community at large to produce the most powerful spreadsheet in the world."
The default size is the same as MS Excel (256x64k). That helps ensure that all the funky xls files out there that depend on those limits work out of the box. However, those values are simple #defines. All coordinates are 32 bits internally. A quick edit and a recompile will change the bounds.
It's not quite ready for prime-time yet, but this is getting closer to being able to code your macros in Perl.
.xsession my friend, this is where this kinda stuff is supposed to go...
True genius is grasping a situation like a peice of fruit, and peircing it just right so that it drains dry.
There are two somewhat related issues to contend with.
1) The file formats are semi documented. We have rough ideas of what OLE2, BIFF[5-8], and escher look like. There are however, lots of abiguities and question marks. As a result we have lots and lots of validation on what get imported. OOo does the same, which is why we can frequently crash MS excel when adding something new to the xls exporter, but still be able to read each others output.
2) The format for VBA is undocumented a far as I know. OOo has a few guesses in place and I've started doing some research on it, but neither of us can even read the vba enough to worry about running macro viri.
3) what scripting capabilities we do have (eg in python, perl, or guile) are strictly sandboxed. We are definitely tending to err on the side of caution rather than functionality.
Correct versions of several functions that are broken in MS Excel. We need to support their incorrect values and the right answers so that imported workbooks stay the same.
More Statistics
More Random distributions
Lots of financial derivative pricers
We're working on it. There is only one remaining technical requirement that is not available under win32, and the new egg menu/toolbar code will fill that slot nicely. Having a win32 build is a high priority. It may not make it in time for 1.2.0 next month but it will go in before the end of the year.
1.1.x has an importer for sxc documents.
.gnumeric exporter for OO one day, but don't see much use in it given that we can read their native format and their xls.
It could be improved, but the heavy lifting is in place and the rest just requires some attention to detail. An exporter will be added eventually.
I'm tempted to write a
Please don't judge all of gnumeric based on the text import in 1.0.x. There have been lots of performance improvements and enhancements there in the development series. The core of gnumeric is easily capable of handling that magnitude of data. Try 1.2.x when it comes out next month (or even 1.1.x if you want to help beta things).
MS Excel is still somewhat faster mainly due to its memory foot print. It was written back in the day and bit bashes things all over the place. Gnumeric pays a penalty for using 32bit addresses rather than bit bashing 18.
If you have something that performs badly please _tell_ us. Our goal is to produce the best damn spreadsheet around. This is still version 1.1, 2.0 (extend) and 3.0 (extinguish) aren't due for a while yet.
We tend to split extension into 2 areas
1) writing functions. Which is supported and documented in python, perl, and guile (and of course compiled languages)
2) scripting. Which is currently unfinished and intentionally mostly undocumented. There are some experimental bindings for python, but we have not had the time to select a solid enough api that we could commit to it. Gnumeric tries to under promise features, and I don't want to whip out some half baked api. The 1.3 development cycle will target scripting and we'll likely wrap the selected api in python, perl and corba initially.
We could use some help on this.
Gnumeric has significantly better statistical routines than MS Excel. Some are home grown, others are from R.
Our charting utilites still have alot of growing to do. Hopefully with the new framework in place now we can start to accumulate features to target something like grace/xmgr