Thanks Vic. Although IMHO the real bug is that I can't actually tell from
the documentation whether this behaviour is abnormal or normal. Could we
fix that first?
For reason's I've already outlined, I'd prefer that if I specify:
<gs:data startRow="2">
<gs:column index="A" name="ID" />
<gs:column index="B" name="Description" />
<gs:column index="C" name="Notes" />
</gs:data>
... then I'd rather the names that I put there 'stick'. Notice that if I do
specify this, then <gs:header row="1"/> is totally useless information -
I've already told you the row where my data begins and the names of my
columns. I genuinely don't understand why we need to supply <gs:header/> at
all.
On the other hand, if I were to supply <gs:header row="1"/> and you were to
derive my column names from that, it's almost insulting that I have to
provide the <gs:column/> information. The only possibly reason for it would
be to 'filter' the columns actually used by the table. In which case, this
would be the more appropriate thing to supply:
<gs:data startRow="1">
<gs:column index="A" /> <!-- name already bound -->
<gs:column index="B" /> <!-- name already bound -->
<!-- column C omitted from this 'table' -->
</gs:data>
Supplying BOTH a <header/> and a 'name' on <gs:column/> only serves to
confuse be and/or the implementors. Can you please clarify the
documentation on this so we can at least know what we should expect?
thanks,
David.
