jamesrcounts/usaddress

OutOfMemoryError parsing some addresses

Closed this issue · 4 comments

lscharen commented

I have been using this library successfully for quite a while, but just deployed to a production environment and the ParseAddress method is causing the .Net Regex class to trigger an OutOfMemoryError. I have not seen this problem during development -- it only is happening in production.

Some sample addresses that trigger the behavior (but the exact address may not matter):

  • 4611 KOLZE AVE, SCHILLER PARK, IL
  • 1654 ILLINOIS ST, DES PLAINES, IL
  • 11 W HUBBARD ST, CHICAGO, IL

Any suggestion of what we can look for that may be triggering this behavior? We have very limited access to the production environment, so just dropping into a debugger is not feasible.

oom

tetsuo13 commented

I've experienced similar problems only in production. Have you tried disabling compiling regular expressions as mentioned in #1?

lscharen commented

@tetsuo13 Yes, I updated to the 0.19 nuget package anddisabled regex compilation and that seems to have worked around the issue.

jamesrcounts commented

Should we close this? Or does someone want to submit a PR to make no-compilation the default?

lscharen commented

Closed. Curious as to the underlying cause, but having a workaround is good enough for me.