Re^5: Bioinformatic task

[erix]

The last time I checked ...

Last time you checked? You didn't check, you took the reported and extremely slow value ("half hour") at face value, instead of comparing your version with a bioperl-version.

In that same thread I showed the difference between bioperl or non-bioperl was much smaller than "half hour" vs "11 seconds".

Now to advocate avoiding a heavy framework for speed is fair enough. But you shouldn't repeat extreme slowness unchecked & unvalidated (it's a bit similar to this slanderous assertion about database - though, of course, that was worse! :) ).

[BrowserUk]

True. I took the OP at face value. Just as I did with you and your timings in the db/sort case.

However, after the I first became aware of the performance problem, I did eventually get (parts of) BioPerl installed on my own (old) machine and verified that the problem existed. I also tried to hunt down the cause.

One possibility that was muted for this was the use of $`, $& or $'. Although I suggested to the OP that he try Devel::SawAmpersand, there was no follow up, so I hacked my way to installing Bio::SeqIO::fasta and all the required support packages, manually. (The full package has never installed for me.) I didn't detect that particular problem, but saw the machinations the module went through (back then; I see quite a lot of it is now commented out), to read a text file, and attributed the (at that point, quite extreme) slowness to that. I also attempted to track the module changes for a while, but that went out the window when I got my current machine.

Long story short, when the OP reports "half an hour" for 200MB, that fit with my prior experiences, and I took him at his word. As your reply in that thread was not to me, I probably never saw it. So, if you are responsible for the performance improvements in the module, I congratulate you.

All of that said, performance was only one of the issues I raised above.