py3: Fix interfaces module for python3

[vklein]

There is still two remaining problems :

• I temporary used sleep(float(0.5)) to fix a doctest but it can't be a good solution. The problem is using sleep(0.5) doesn't pause the thread 500 ms with python3 :

sage: time sleep(0.5)
CPU times: user 48 µs, sys: 0 ns, total: 48 µs
Wall time: 52.2 µs
sage: time sleep(1)
CPU times: user 468 µs, sys: 8 µs, total: 476 µs
Wall time: 1 s

Does somebody knows if there is a ticket related to this problem or if i should open a new one ?

• In pexpect.py the doctest

  sage: w = walltime(t); w > 0.4 and w < 10
  True
  

fails in python3 because the walltime is shorter than 0.4. I don't know why it should be considered as an error or if it is expected with python3 for some reason.

[vklein]

New commits:

​91bdf2a

Trac #26314 : module interface, fix code and doctest for...

[chapoton]

see #26311 for the sleep problem

[vklein]

Thanks !

[vklein]

This ticket don't manage the walltime case.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​8f35e93

Trac #26314: Remove some unused import and variable.

[vklein]

Remove some unused variable to make pyflakes more happy (AsciiArtString is needed by macaulay2.py).

[chapoton]

@embray and @jdemeyer, any opinion on the simple changes proposed here ?

[vklein]

... or an opinion about the "walltime" test ...

[embray]

For cases like:

src/sage/interfaces/sagespawn.pyx

@@ -145 +145 @@
             sage: from sage.interfaces.sagespawn import SageSpawn
             sage: E = SageSpawn("sh", ["-c", "echo hello world"])
             sage: _ = E.expect_peek("w")
-            sage: E.read()
+            sage: E.read() # py2
             'hello world\r\n'
+            sage: E.read() # py3
+            b'hello world\r\n'

where Python 3 returns bytes, instead of making two separate cases, in the past I have done something like print(E.read().decode('ascii')). This will have the same result in both cases.

I wonder if Sage shouldn't include its own sleep() built-in that isn't broken :/

[vklein]

Replying to embray:

where Python 3 returns bytes, instead of making two separate cases, in the past I have done something like print(E.read().decode('ascii')). This will have the same result in both cases.

I know, i hesitated between the two syntax and, as it is an EXAMPLES::, have think that the # py2-# py3 syntax might be clearer for the end-user.

I don't have a strong opinion on it, tell me what you prefer. If the past cases you talk about are also EXAMPLES:: we should stick with the decode('ascii') solution for consistency.

[embray]

Yes, I think consistency is best. There are a few other cases that also, rather than relying on the repr, just do like

sage: E.read() == b'Hello\r\n'
True

or whatever. This will also work the same on Python 2 and 3.

I have thought of adding a helper to the test framework to deal with b-prefixed strings as it does with u-strings, but it turns out they are relatively uncommon in the majority of the test suite.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​4417bbb

Trac #26314 : replace '# py2 - # py3' with ...

[vklein]

# py2-# py3 cases replaced using decode(ascii) syntax.

[chapoton]

probably you do not need the initial u that you add to some doctests.

[embray]

Replying to chapoton:

probably you do not need the initial u that you add to some doctests.

I think it is needed because on Python 2 .decode(...) will return a unicode string literal, and the doctest parser does treat this as relevant on Python 2. The only thing it does w.r.t. u-prefixes is on Python 3 it strips them from the expected output.

I usually avoid the issue by using print(...) instead, but i think this is fine as-is.

[vklein]

It is as Eric says, these tests as they are won't work on py2 without the initial u

[chapoton]

ok, let it be

[embray]

This should be re-targeted for 8.5.