Confirm: compiled re(gexps) are thread safe?

[Johan Ovlinger]

Subject says it all, really.

Can someone please confirm that I can safely do something like the below,
without needing a thread-local compiled regexp? (not run, so probably rife
with syntax errors, but you get the drift)

global_var = re.compile( "foo" )

for str in manystrings:
def domatch(str):
if global_var.search(str):
print "yahooo!"
threading.thread(target = domatch, args = [str]).start()

Alternately, what is the overhead in compilation? Neglible?

Thanks,

Johan

 

[Skip Montanaro]

Johan> Subject says it all, really.

Yes, searching using a compiled regular expression is thread-safe.

Skip

 

[Fredrik Lundh]

Can someone please confirm that I can safely do something like the below,
without needing a thread-local compiled regexp?
yes.

global_var = re.compile( "foo" )

for str in manystrings:
def domatch(str):
if global_var.search(str):
print "yahooo!"
threading.thread(target = domatch, args = [str]).start()

Alternately, what is the overhead in compilation? Neglible?

the compiler uses a cache, so even if you call compile inside each thread,
all threads end up getting the same pattern object.

on the other hand, make sure you google for "global interpreter lock" before
you spend too much time implementing parallell searches...

</F>

 

[Johan Ovlinger]

Johan> Subject says it all, really.

Yes, searching using a compiled regular expression is thread-safe.

Skip

Great.

Thanks