Urllib2 uses both hashlib and sockets. Neither work right in IronPython, and I can’t seem to get them working with FePy either. I replaced “os.access” calls with its partial equivalent “os.path.exists” (which pulls from ntpath.py on windows). That worked enough to validate that sockets was an issue. Even pointing at FePy’s libs, I can’t reference “socket._fileobject.” I can see the code, and FePy talks about fixing this, but I’m clueless as to what I’m doing wrong.

Basically, I’m not sure IronPython support in TileCache is worth worrying about yet. In 6 months, all these issues should be resolved, and things “just work”.

There must be some method of determining whether a file exists: if you wanted to, you could create a subclass of DiskCache (IronPythonDiskCache) and replace the __init__, __set__ and __delete__ methods. (If os.makedir is also missing, you might need to replace the attemptLock function too.)

Also, I’m not sure which 44k of code you’re talking about, but the TileCache Python files (minus examples, docs, etc.) are only 1200:

disciplina:~/tilecache-1.8.1/TileCache crschmidt$ ls
Cache.py Client.py Layer.py Service.py __init__.py
disciplina:~/tilecache-1.8.1/TileCache crschmidt$ wc -l *
154 Cache.py
115 Client.py
418 Layer.py
526 Service.py
5 __init__.py
1218 total

So it’s even worse than you suspected

Out of curiosity, where is _hashlib getting imported from? Is it something we can drop out of TileCache in the main branch? I’d like to TC and make it work on IronPython out of the box if possible, rather than having a forked project — if there’s ways we can test for IronPython and change some of our behavior, I’d love to see them.

http://docs.python.org/lib/os-file-dir.html

Interesting exploration by the way.

Fuzzyman