[{"user_id": 8649, "stars": [], "topic_id": 14067, "date_created": 1300628760.482331, "message": "I had a nice chat with Wes McKinney (author of Pandas http://code.google.com/p/pandas/) yesterday and he was bemoaning the lack of visibility of Python's success in the financial space.  I ventured that the closed-door nature of finance really hurts everyone in that space, because they all repeat each others' mistakes.  Most firms struggle with the same questions and challenges, and the lack of knowledge sharing makes everyone less efficient.", "group_id": 6727, "id": 392028}, {"user_id": 8649, "stars": [], "topic_id": 14067, "date_created": 1300629080.961201, "message": "First (and most obviously) it would be great to actually have a resource/forum/mailing-list where folks can actually discuss their needs.  Building on that knowledge, we could build out an actual package a la scipy for quantitative finance.  I think a Cython wrapper for QuantLib would be pretty cool, but it needs to somehow play nicely with numpy/scipy as well.", "group_id": 6727, "id": 392034}, {"user_id": 10411, "stars": [], "topic_id": 14067, "date_created": 1300652207.285398, "message": "@pwang That probably missed the point.   I agree.   The success of Python in the finacial space is far, far, far understated.", "group_id": 6727, "id": 393599}, {"user_id": 10411, "stars": [], "topic_id": 14067, "date_created": 1300652140.5293219, "message": "@pwang QuantLib claims they have a Python wrapper already.  The benefit of a Cython wrapper would be...?  (You could call it from other Cython code nicely?)", "group_id": 6727, "id": 393591}, {"user_id": 8649, "stars": [{"date_created": 1300687580.828105, "user_id": 10411}, {"date_created": 1302941091.332499, "user_id": 23030}], "topic_id": 14067, "date_created": 1300670689.547972, "message": "@scopatz They have a SWIG wrapper now.  <cue organ music>", "group_id": 6727, "id": 394662}, {"user_id": 8649, "stars": [], "topic_id": 14067, "date_created": 1300671140.536927, "message": "A Cython wrapped qlib means you can optimize other routines more easily in cython.  Having a Cythonized scipy also means that we can build a lot of interop at the C/Cython level, which is very advantageous.", "group_id": 6727, "id": 394685}, {"user_id": 10411, "stars": [], "topic_id": 14067, "date_created": 1300687565.56073, "message": "@pwang I get the interop wins (esp with SciPy).  But rewriting some else's wrappers seems like a huge headache. Even bigger than writing my own ;)", "group_id": 6727, "id": 395608}, {"user_id": 21520, "stars": [], "topic_id": 14067, "date_created": 1300728659.8721581, "message": "Discussing with some folks about the current state of the quantlib wrappers, it seems there is still some work to be done to expose some internal features like path generators, etc. I started some tests last year but had issues with namespace support in Cython. I think it is solved now. I should give it a second shot.", "group_id": 6727, "id": 400430}, {"user_id": 21157, "stars": [], "topic_id": 14067, "date_created": 1300749310.659214, "message": "There was a thread just yesterday on the numpy mailing list about wrappers around GAUSS for doing economic computation in python; not sure how relevant this is since I don't know much about the financial python world myself, but in case it's helpful: http://thread.gmane.org/gmane.comp.python.numeric.general/43138", "group_id": 6727, "id": 403497}, {"user_id": 8649, "stars": [], "topic_id": 14067, "date_created": 1300756004.231539, "message": "@dpinte Yes, there is namespace support in Cython.  It seems to work fine for the kiva wrappers I was working on.", "group_id": 6727, "id": 404160}, {"user_id": 21520, "stars": [], "topic_id": 14067, "date_created": 1300815713.639801, "message": "@pwang Things do seem to work fine. I'll go on with my tests and see if I can post a useful stuff soon", "group_id": 6727, "id": 409403}, {"user_id": 11207, "stars": [], "topic_id": 14067, "date_created": 1301027547.6934731, "message": "I like the idea of a basic set of tools ala a *nicely* wrapped quantlib for Python.  This could really push the exposure of Python in Finance.", "group_id": 6727, "id": 435544}, {"user_id": 11207, "stars": [], "topic_id": 14067, "date_created": 1301027637.787931, "message": "Like others have mentioned, the interop with NumPy and SciPy could make the package very useful.", "group_id": 6727, "id": 435563}, {"user_id": 11207, "stars": [], "topic_id": 14067, "date_created": 1301027605.3159621, "message": "It would be useful to think a bit about the interface, though.", "group_id": 6727, "id": 435552}, {"user_id": 21711, "stars": [], "topic_id": 14067, "date_created": 1301072422.552922, "message": "I tried using quantlib for option pricing once, but when I got really strange results when trying to use their models which incorporate discrete dividends", "group_id": 6727, "id": 440129}, {"user_id": 10411, "stars": [], "topic_id": 14067, "date_created": 1301284447.9234819, "message": "@pwang Agreed.  We clearly lack a first cut right now, though.  Maybe the current SWIG wrapper counts, but since not many people use it and no one seems bent on improving it...", "group_id": 6727, "id": 454692}, {"user_id": 8649, "stars": [], "topic_id": 14067, "date_created": 1301283228.1433761, "message": "I'm quite certain that any library that succeeds in this area will require a lot of iteration and input from practitioners in the field.  However, I feel like if people had a basic architecture for pricing/pricers, tree models, distributions, etc., that would be at least a starting point.  Furthermore, it could also define a first cut at interfaces for people to build data importers/exporters and visualization tools and UIs.", "group_id": 6727, "id": 454548}, {"user_id": 10411, "stars": [], "topic_id": 14067, "date_created": 1301284473.1540711, "message": "Better wrappers (read: cython) would certainly help.", "group_id": 6727, "id": 454694}, {"user_id": 21520, "stars": [{"date_created": 1302020455.3636279, "user_id": 10411}], "topic_id": 14067, "date_created": 1301989093.7958331, "message": "first feedback from the quantlib tests : 7 times faster than the SWIG ones and much more pythonic ;-)", "group_id": 6727, "id": 537871}, {"user_id": 10411, "stars": [], "topic_id": 14067, "date_created": 1302020484.095866, "message": "@dpinte Do you have it posted somewhere?", "group_id": 6727, "id": 542995}, {"user_id": 8649, "stars": [], "topic_id": 14067, "date_created": 1302051343.1584921, "message": "@dpinte Nice!  7x, are you serious?  I wonder what the hell the SWIG wrappers were doing!", "group_id": 6727, "id": 551484}, {"user_id": 10454, "stars": [{"date_created": 1302941131.013087, "user_id": 23030}], "topic_id": 14067, "date_created": 1302098245.5047021, "message": "Python function call overhead?", "group_id": 6727, "id": 559486}]