MAINT: Fixup and remove bundled dependencies #1234
Conversation
HaoZeke
force-pushed
the
noExternal
branch
2 times, most recently
from
898c148 to
3f39e6f
Compare
|
Will debug soon. |
HaoZeke
force-pushed
the
noExternal
branch
2 times, most recently
from
ab99396 to
689edee
Compare
|
This could use a refresh |
asv/benchmark.py
Outdated
| return | ||
|
|
||
| try: | ||
| subprocess.check_output(['python', "-mpip", "install", 'pympler']) |
There was a problem hiding this comment.
This works with all three environment backends, conda, mamba (as of #1238) and virtualenv but is syntactically less than ideal. A better approach would be to modify the constraints of the environment class which is the base of all the backends.. Though YAGNI probably.
|
After the move from setup.cfg to pyproject.toml this now has conflicts. |
MAINT: Kill extern in favor of pympler and pyjson5 Revert "MAINT: Kill extern in favor of pyjson5" This reverts commit 02a2b170910f55072a008cc7d52da55e0b894b32. MAINT: Fixup MAINT: Use pyjson5 correctly
HaoZeke
force-pushed
the
noExternal
branch
3 times, most recently
from
0f5aa42 to
bbd8f8a
Compare
HaoZeke
force-pushed
the
noExternal
branch
10 times, most recently
from
b8bb50a to
068546d
Compare
HaoZeke
force-pushed
the
noExternal
branch
2 times, most recently
from
6c443ea to
bb87d6d
Compare
HaoZeke
force-pushed
the
noExternal
branch
2 times, most recently
from
9d85a5f to
b2c6f9a
Compare
| return | ||
|
|
||
| def import_asizeof(): | ||
| """Import asizeof, searching system Pythons in PATH.""" |
There was a problem hiding this comment.
Why do we need all this? If the import fails, can't we just fail the run?
There was a problem hiding this comment.
The problem here is that the import fails where it shouldn't, e.g. on the CI. This seems to be because of the process isolation (each ASV run is in a different subprocess, which doesn't include asv and on CI cannot use pip either). Perhaps, now that I'm thinking about it, it would be cleaner to instead import asv itself, which would automatically bring in pympler and is discussed in #908
There was a problem hiding this comment.
If I went with that approach then the Import Mechanism would instead load ASV rather than asizeof, but I'm not sure how to simplify this further without breaking the subprocess approach.
HaoZeke
force-pushed
the
noExternal
branch
8 times, most recently
from
0791a61 to
4ea96eb
Compare
|
I am fine to merge this as-is and then to iterate on solving the import problem, what do you think? |
Sure that sounds great. |
This supersedes #1225 and #1211.
Also closes #1139.
cc @mattip