-1 on committing any patches without unit tests anymore. It destabilizes code too much. Even to scratchpad because then the code never moves out of scratchpad. We learned this lesson already. It doesn't matter how good or bad or how much we want it...no unit tests + doco...it stays as patches in bugzilla.

[EMAIL PROTECTED] wrote:
[Moving to the list, since bugzilla is bad for conversation]

Comments inline.

**** As a result, you may have to apply the   ****
**** patch to /src/scratchpad/src instead of  ****
**** the root folder                          ****


That's ok. For the initial bit, can you just zip/tgz the directory. Provide
patches after the initial additions have been comitted. I am waiting to commit
till the dummy method javadocs are removed (maybe an awk script might be
quick?)



1. Ptg and Eval: Delegating or extending? <snip> But I guess extending from Ptg could also work...


Just to clarify, by extend, i meant have a separate class that interits from the
Ptgs.. ie, `class AddEval extends AddPtg` .. but as I said, just a suggestion,
you'll know best.


2. Functions in one class or one class per function?

<snip>

Yeah, sure.

3. Testing...
Unit testing would be a big effort. Almost as big as writing individual
functions, a distributed effort would be great!


Certainly, probably bigger. So once again, this is a call for everybody
interested in this functionality to write unit tests for it. Its easy!


4. My eclipse...

<snip>

Please try and remove the extraneous comments, i dont want to put that in CVS.
I'll commit it as soon as that's done, I'm happy with everything else.


5. Under contruction:
The work on core eval classes is not yet "complete". I think there is need for a
BlankEval class to handle empty cells - which are currently being handled as
StringEval("").


Yeah, sure, thats understood, and no issues. But your 'empty cell' comment made
me think, the function implementations should probably follow Excel semantics..
for eg, in the average function, empty cells do not add to the number of
elements (ie, the denominator). Also, these semantics might be different for
different versions of excel. How do we handle that.


A final thoguht on error handling. Your error handling is modelled after OO.o.
In excel however, errors are a more limited set.. #REF, #VALUE, etc. Do we need
to think about how to handle/map those. Maybe it doesnt matter?


Once again, to everybody out there, this is a big and important piece of work
that can certainly do with more evaluation/comments/unit tests. Thanks!


Regards
-
Avik



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
Mailing List:    http://jakarta.apache.org/site/mail2.html#poi
The Apache Jakarta POI Project: http://jakarta.apache.org/poi/
.



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
Mailing List:    http://jakarta.apache.org/site/mail2.html#poi
The Apache Jakarta POI Project: http://jakarta.apache.org/poi/



Reply via email to