bruns marked an inline comment as done.
bruns added inline comments.


> poboiko wrote in filecontentindexer.cpp:91
> OK, we can simply count how many times we got `finishedIndexingFile`, and 
> just go to the corresponding position in the batch.
> It's just the binary search here does look a bit unnecessary to me...

Counting would give a good hint which one failed, but still no guarantee. Files 
may be added or removed during the indexer run, so the position is only 
approximate. The indexer may also have sent a "finished" message, and crash 
afterwards in a destructor call.

Doing a binary search is straight forward and avoids any dependencies or 
assumptions about other parts of the code.

Also, this code should be only temporary anyway - if the extractor is run in a 
separate process which only receives the file using a readonly file descriptor 
(for sandboxing) and passes back the result, the problematic documents id is 
known by the parent process.

  R293 Baloo


To: bruns, #baloo, #frameworks, poboiko, ngraham
Cc: broulik, apol, kde-frameworks-devel, ashaposhnikov, michaelh, astippich, 
spoorun, ngraham, bruns, abrahams

Reply via email to