A simple search for $ and replace with (say) Z2Z with Regular Expressions enabled will only work on small documents.
In all mt RTF documents I use a series of macros to format and tidy the documents to a standard output. The above search and replace is only one of many simple steps which remove indents, join together paras where the line is broken unintentionally etc.
On a small document the process works fine. Once there is say 300 A4 pages, the process never fully ends.
In the 3x series, this was a rapid time to completion, regardless of size of document.
Changing allocated memory via options has not had any appreciable effect. I do however suspect a memory leak at this time as I see resources used continuously increasing during the process. Force quit is required to resume.
Running the same process in the latest OpenOfiice (4.1) shows that the Search and Replace completes very effectively.
Thank you for reporting the bug. Is it possible for you to supply us with a sample document so we can run tests on it to determine where the problem arises.
Documents sent by email to Jay Philips
Thanks Oldbwl for the document. The problem seems to have arisen starting in the 4.1 branch.
Performance Tests :-
4.0.6 : 2 secs
4.4.0 : > 5 minutes (canceled when it reached there)
Build ID: 4774b2f0e6a0e895d1e63256147eaff59efd0eda
TinderBox: Linux-rpm_deb-x86@45-TDF, Branch:libreoffice-4-4, Time: 2014-12-29_17:48:44
Hmm, search & replace over a few hundred pages (how many cells and with what content ?) - should be quick. Then again, an ICU upgrade seems to have been behind a lot of slow-downs in recent LibreOffice/calc bits.
It'd be really nice to have either a document that displays this problem; or a series of simple instructions to build a document that reproduces this. Also - Eike is rather an expert in ICU-ness and calc, so CC'ing him =)
(In reply to Michael Meeks from comment #5)
> Hmm, search & replace over a few hundred pages (how many cells and with what
> content ?) - should be quick. Then again, an ICU upgrade seems to have been
> behind a lot of slow-downs in recent LibreOffice/calc bits.
The supplied document is an rtf with 181 pages loaded in Writer, which the search replaced 2011 instances of regexp '$' with 'Z2Z'.
> It'd be really nice to have either a document that displays this problem; or
> a series of simple instructions to build a document that reproduces this.
> Also - Eike is rather an expert in ICU-ness and calc, so CC'ing him =)
The document has a novel/story structure, meaning simple paragraph text with quotes when people are speaking.
@Oldbwl: Is it okay for me to share the document with the developers for testing?
@JayPhillips: As long as it is not published online yes,
Almost certainly a duplicate - and 80715 is better isolated to a single commit =)
*** This bug has been marked as a duplicate of bug 80715 ***
Migrating Whiteboard tags to Keywords: (perf bibisectRequest)