On Tue, 19 Feb 2008, Josh Micich wrote:
Around line 160 there is a reverse iteration loop over the shared
formula records. The fix to 44449 was to limit the content of the
sharedFormulas list (around line 140). Without that fix, this reverse
loop causes a version of the bug which is much easier to reproduce:
Make a spreadsheet with a 2 distinct shared formulas - one in
Sheet1!A1:A8 and another in Sheet2:A1:A8. Then use POI to read the cell
formula for Sheet1!A2. It will incorrectly give the shared formula from
the second sheet.
Is there some record in the steam we can use to spot when we move from one
sheet to the next one?
And now to lay some blame ...
svn blame
src/java/org/apache/poi/hssf/record/aggregates/ValueRecordsAggregate.java
...
610216 nick // Traverse the list of shared formulas in
610216 nick // reverse order, and try to find
the correct one
610216 nick // for us
426082 jheight boolean found = false;
426082 jheight for (int i=sharedFormulas.size()-1;i>=0;i--) {
I think I was just tidying up the comment when I tweaked it...
Changing the loop back to the forward direction does not break any
junits. I put a comment in the code to that effect.
My best guess is it's like that for repeatedly saved files (so there is
an old version and a new version of the record), or in the case of
overlapping shared formulas (eg A1->A3 then A1->B6 or something like
that). Does that sound likely to anyone else?
If we can find some record / condition that'll let us break the shared
records up into one group per sheet, that'd seem to me to be the ideal
fix.
Nick
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
