COMPAT: 3.6 #14679
Comments
jreback
added
Compat
jreback
added a commit
to jreback/pandas
that referenced
this issue
Nov 17, 2016
jreback
added a commit
to jreback/pandas
that referenced
this issue
Nov 17, 2016
jreback
added a commit
that referenced
this issue
Nov 18, 2016
partial on #14679 Author: Jeff Reback <[email protected]> Closes #14681 from jreback/warnings and squashes the following commits: 56e7b1b [Jeff Reback] COMPAT: remove some deprecation warnings in 3.6
|
some of these pickle issues are due to an apparent change in how the pickle methods, specifically |
jreback
added a commit
to jreback/pandas
that referenced
this issue
Nov 18, 2016
jreback
added a commit
to jreback/pandas
that referenced
this issue
Nov 18, 2016
jreback
added a commit
to jreback/pandas
that referenced
this issue
Nov 18, 2016
jreback
added a commit
to jreback/pandas
that referenced
this issue
Nov 18, 2016
|
cc @llllllllll |
|
I will look into this tomorrow. |
jreback
added a commit
that referenced
this issue
Nov 18, 2016
xref #14679 Author: Jeff Reback <[email protected]> Closes #14685 from jreback/offsets and squashes the following commits: 8ad212c [Jeff Reback] BUG: fix pickling of Custom offsets in 3.6
|
@jorisvandenbossche not sure why this fails on 3.6 |
|
https://travis-ci.org/jreback/pandas/jobs/176992785 is latest build. |
|
@jreback regarding sql error: not directly an idea. I can reproduce it by running the specific test, but not with a standalone snippet: runs fine |
jreback
pushed a commit
that referenced
this issue
Nov 22, 2016
In python 3.6, the CALL_FUNCTION handling was updated. One change is that when calling a C function from a python function python now counts the reference owned by the argument tuple. This means that move was always seeing objects with two references instead of the expected one. Python 3.6 also removed a copy in the argument tuple when *unpacking functions. This means that if a user does: tuple = (create_string(),) move_into_mutable_buffer(*tuple) where create_string() creates a string object with one reference then we will fail to raise a BadMove even though the user could later retrieve that string with tuple[0]. There is no way to detect this case so this patch adds a warning to the docstring advising against star unpacking. xref: #14679 I played around with removing the extra reference that was added in 3.6 but it looks like playing with borrowed refs everywhere will be a bit tricky. This change should clear things up for 3.6 while continuing to work for older versions. In 3.6 you __could__ get a shared mutable string from this but you need to try pretty hard for it. Author: Joe Jevnik <[email protected]> Closes #14695 from llllllllll/move-3.6-compat and squashes the following commits: d1e8b1b [Joe Jevnik] BUG: Fix move_into_mutable_buffer for python 3.6.
|
Jeff, apart from the sql problem for which I opened another issue, are all of the above errors solved in the meantime? Or are there still ones to look at? |
Closed
|
nope that's it |
jorisvandenbossche
pushed a commit
to jorisvandenbossche/pandas
that referenced
this issue
Dec 14, 2016
partial on pandas-dev#14679 Author: Jeff Reback <[email protected]> Closes pandas-dev#14681 from jreback/warnings and squashes the following commits: 56e7b1b [Jeff Reback] COMPAT: remove some deprecation warnings in 3.6 (cherry picked from commit 748000d) Fixed conflicts in: pandas/io/tests/test_clipboard.py
jorisvandenbossche
pushed a commit
to jorisvandenbossche/pandas
that referenced
this issue
Dec 14, 2016
xref pandas-dev#14679 Author: Jeff Reback <[email protected]> Closes pandas-dev#14685 from jreback/offsets and squashes the following commits: 8ad212c [Jeff Reback] BUG: fix pickling of Custom offsets in 3.6 (cherry picked from commit bec5bdb)
jorisvandenbossche
pushed a commit
to jorisvandenbossche/pandas
that referenced
this issue
Dec 14, 2016
…Timestamp input TST: fix unordable error message xref pandas-dev#14679 TST: handle unorderable exceptions in indexing closes pandas-dev#14684 (cherry picked from commit b6ffd89)
jorisvandenbossche
pushed a commit
to jorisvandenbossche/pandas
that referenced
this issue
Dec 14, 2016
…hon 3.6. In python 3.6, the CALL_FUNCTION handling was updated. One change is that when calling a C function from a python function python now counts the reference owned by the argument tuple. This means that move was always seeing objects with two references instead of the expected one. Python 3.6 also removed a copy in the argument tuple when *unpacking functions. This means that if a user does: tuple = (create_string(),) move_into_mutable_buffer(*tuple) where create_string() creates a string object with one reference then we will fail to raise a BadMove even though the user could later retrieve that string with tuple[0]. There is no way to detect this case so this patch adds a warning to the docstring advising against star unpacking. xref: pandas-dev#14679 I played around with removing the extra reference that was added in 3.6 but it looks like playing with borrowed refs everywhere will be a bit tricky. This change should clear things up for 3.6 while continuing to work for older versions. In 3.6 you __could__ get a shared mutable string from this but you need to try pretty hard for it. Author: Joe Jevnik <[email protected]> Closes pandas-dev#14695 from llllllllll/move-3.6-compat and squashes the following commits: d1e8b1b [Joe Jevnik] BUG: Fix move_into_mutable_buffer for python 3.6. (cherry picked from commit f609640)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
after #14678
Bunch of Deprecation Warnings
only a small number of 3.6 failures remain.
The text was updated successfully, but these errors were encountered: