Basic exercise of ipython [subcommand] -h and help-all
#4125
Conversation
|
The decorators seem a bit on the magical side given the need for them, but that's a matter of preference. Other than that, +1. |
|
I agree - I just didn't want to do a huge amount of copy/paste in case anything should change. Do you have a better way to generate a bunch of tests that differ by a simple string key? |
|
I think I'd just use |
|
decorators removed, tests generated now from a text-editor macro instead of at runtime. |
|
LGTM |
it has `config=True` traits Currently in master `ipython console -h` will fail because of this. It doesn't affect 1.0, I'm not sure why.
should help catch when we break these things.
prevented `nbconvert -h` from working
help-all is a superset of `-h`, so no need to do both, since these tests are slow.
|
I expressed concern in person that this slows down our test suite. In particular, here are the running times on my laptop with an SSD for this PR: Ran 14 test groups in 172.081s whereas master on cold cache, and warmed up runs in about 90 seconds. This may or may not be worth updating, since it'll become less of an issue when we re-factor and parallelize the test suite. We also currently have no way of marking tests as "slow" and to skip them unless the (these times were before @minrk updated to only run |
|
actually, an update to my earlier comment - i hadn't updated IPython master on my laptop for a week, and ipython master test suite now runs in 160 seconds :( Min's looking into this right now |
|
Almost all of the new slowness is in new kernel tests introduced in the last week or so, which spin up new kernels, some more than necessary (it can take up to three seconds to start and stop a kernel subprocess), not this PR. I will investigate reusing kernel instances for more tests, which should help. But that's for another PR. |
|
merging soon unless objections - traceback on nbconvert help output is bugging me. |
|
lemme just test it locally one more time |
|
I'm doing the same |
|
looks good, i haven't pushed the green button for a while, so I'm gonna go for it! |
Basic exercise of `ipython [subcommand] -h` and help-all
|
lemme see if the backport script still works, now... |
|
will need to be careful with the backport, because the Transformer/Preprocessor change will want to be excluded. |
|
hmm, so an hour later, can I ask you for an assist, @minrk on the backport? I've edited the patch down to just two error which don't apply, but don't know how to proceed :\ error: patch failed: IPython/kernel/tests/test_kernel.py:250 |
|
sure, I'll look at it. |
|
worked with this patch |
…lp-all for various subcommands. Prompted by the fact that `ip qtconsole -h` and `ip nbconvert -h` are both broken in master. This PR also fixes both. Should be backported to 1.1
Basic exercise of `ipython [subcommand] -h` and help-all
for various subcommands.
Prompted by the fact that
ip qtconsole -handip nbconvert -hare both broken in master.This PR also fixes both.
Should be backported to 1.1