Quote:
Originally Posted by kiwo22
This regression-test-file folder contains too few files... It only has "cash" (no SNG or MTT); Stars folder, which is supposed to be the biggest one, only has several "Draw" files with no Holdem or Stud etc.
I'm not sure what you are looking at but that is definitely not all of them.
http://trac.assembla.com/fpdboz/brow...ion-test-files
Quote:
Originally Posted by kiwo22
Is "a public repository of hand histories for regression testing" that you mentioned something different? If yes, I'm willing to get it for sure ;-)
I meant that if we are both working in parsing projects, then pooling our resources to produce a suite of files that properly exercise all hand history types is something i'm willing to contribute too (and have already
We dont have a database, we have a suite of files with (usually) 1 hand history in each that exercises one or more 'interesting' situations that occur. eg. "All-in pre-plop with 3 callers" or when FTP changed their date format to something different for only 3 days in July 2010
Quote:
Originally Posted by kiwo22
Also, you could easily obfuscate the names of players in files and then share with community. Just a suggestion
Thanks for your help and quick reply
Firstly, i'd never publish more than a single hand history from a player without their express permission. It is hard enough getting people to donate hand histories without are record of blowing that trust.
Second, obfuscating a hh file isn't that simple. We have a tool to assist, but in reality it still needs a manual review. At least one hh file type is sensitive to whitespaces. Dates are important in many cases (and changing the date by a day or two isn't enough to prevent reverse engineering), as are tournament numbers and the format of certain peoples names.
All the problems are solvable, but by the time you have perfect obfuscation, you have a perfect parser. If you have a perfect parser you no longer have a need for a regression test suite and obfuscation.
Sorrow