BUG: strengthen typing in get_c_string, fix StringHashTable segfault #30419
Conversation
| buf = PyUnicode_AsUTF8AndSize(py_string, length) | ||
| else: | ||
| PyBytes_AsStringAndSize(py_string, <char**>&buf, length) | ||
| buf = PyUnicode_AsUTF8AndSize(py_string, length) |
There was a problem hiding this comment.
Hmm interesting - so I guess this helper function existed in the first place for 2/3 compat? If this is all it is reduced to why not get rid of this function altogether and just replace with PyUnicode_AsUTF8AndSize?
There was a problem hiding this comment.
so I guess this helper function existed in the first place for 2/3
thats my guess, yah.
just replace with PyUnicode_AsUTF8AndSize?
for now at least because we need to figure out what to do about the segfault thing; the imlpementation here might end up changing
|
lgtm. not sure why the 36 build is failing though. |
Looks like we're passing |
|
So it isnt a numpy-version problem, just a slow test that is only getting run in that build. We're passing a np.str_ object instead of just a str. This is tough to check for because |
|
lgtm. |
|
thanks @jbrockmendel |
Start with util.pxd, we can tighten the arg from "object" to "str" and simplify it a bit. Tracking down the places where get_c_string is used, the main one is in StringHashTable, where in
set_itemit is currently used incorrectly. The test added by this PR segfaults in master.@TomAugspurger it looks like StringHashTable has relatively light testing. Should we be using it for StringArray?