Re: [Libreoffice] [PATCH] new BUG 36594

well, yes I forgot about that query even may change in translate 
statement, so to store the positions is not always a good solution, you 
are right. Maybe we could simply put the comments after the query 
string, still the user is able to store comments, and the query is ok, 
however the original position is lost, but the order of comments is 
keeped, so I guess it's not such a bad thing. But in this case to use a 
vector is needless, just a simply OUString is enough to store the 
comments. And yes, my problem was about this case, but somehow my former 
patch worked before, is it possible that translate statement has just 
been modified? I guess you should comment out some lines in my patch, 
and put all the comments after the query.

Gabor

2011. 09. 02. 10:09 keltezéssel, Eike Rathke írta:
> Hi Jenei,
>
> On Thursday, 2011-09-01 22:08:33 +0200, Jenei Gábor wrote:
>
> > LibreOffice Base if you turn on running SQL query in SQL edit view
> > it will pass the SQL query directly to HSQL without any type of
> > modification, or check, so if you push this button it should work
> > totally as a HSQL prompt.
> Played a bit with those, while delComment() in parseTree() seems to
> work, the getComment()/concatComment() in translateStatement() does not
> because
>
>              sTranslatedStmt = m_xComposer->getComposedQuery();
>
> never includes LFs in its normalized return, so the following
>
>              sTranslatedStmt = concatComment( sTranslatedStmt, 
> sComments);
>
> doesnt do anything. This may be what you experienced earlier as well
> when you wondered that something was eating the LFs. With the positional
> approach not searching for LFs it may have worked accidentally if
> original query and composed query were identical except one trailing
> comment.
>
> Without messing with the parser itself (and even then we wouldn't know
> if following stages would handle the result) I don't see a clean way to
> solve this, so I suggest to keep delComment() in parseTree() and discard
> getComment()/concatComment(), unfortunately that will lose comments at
> some (unimportant?) stage, but the query form seems to remember the
> original query at least.
>
> If no objection, I'll commit this evening.
>
>    Eike