Implement np.asarray()

[stuartarchibald]

This implements np.asarray() and adds tests.

[stuartarchibald]

Seems like a test that was relying on np.asarray being unsupported is now failing, because this adds support!

[sklam]

Replace types.Float and types.Integer with types.Number so that complex dtype also work.

[stuartarchibald]

done.

[codecov-io]

Codecov Report

Merging #3598 into master will decrease coverage by 0.18%.
The diff coverage is 90.14%.

@@            Coverage Diff            @@
##           master   #3598      +/-   ##
=========================================
- Coverage   80.68%   80.5%   -0.19%     
=========================================
  Files         393     395       +2     
  Lines       80485   81179     +694     
  Branches     9164    9240      +76     
=========================================
+ Hits        64942   65351     +409     
- Misses      14127   14399     +272     
- Partials     1416    1429      +13

[sklam]

ty is frozen as a compile-time constant. the first dt_conv will always be used afterwards

[stuartarchibald]

ah yes, thanks, that explains some of the weirdness

[sklam]

I just realized that my explanation is wrong. The cache is per signature. A different a should not reach into the same cache.

[stuartarchibald]

I think the problem here is due to this #3612

[stuartarchibald]

b6e8f5a has some more tests for different scalar types, merging #3612 into this branch permits them to pass.

[stuartarchibald]

On merge this #3628 can be closed.

[seibert]

This is currently failing two unit tests:

2018-12-31T21:45:49.9986907Z ======================================================================
2018-12-31T21:45:49.9987155Z FAIL: test_asarray (numba.tests.test_np_functions.TestNPFunctions)
2018-12-31T21:45:49.9988076Z ----------------------------------------------------------------------
2018-12-31T21:45:49.9988372Z Traceback (most recent call last):
2018-12-31T21:45:49.9988591Z   File "/home/vsts/work/1/s/numba/tests/test_np_functions.py", line 1604, in test_asarray
2018-12-31T21:45:49.9988808Z     _check(params)
2018-12-31T21:45:49.9989007Z   File "/home/vsts/work/1/s/numba/tests/test_np_functions.py", line 596, in _check_output
2018-12-31T21:45:49.9989216Z     self.assertPreciseEqual(expected, got, abs_tol=abs_tol)
2018-12-31T21:45:49.9989413Z   File "/home/vsts/work/1/s/numba/tests/support.py", line 263, in assertPreciseEqual
2018-12-31T21:45:49.9989609Z     self.fail("when comparing %s and %s: %s" % (first, second, failure_msg))
2018-12-31T21:45:49.9990587Z AssertionError: when comparing 1.2 and (1.2+0j): dtype('float64') != dtype('complex128')
2018-12-31T21:45:49.9990942Z 
2018-12-31T21:45:49.9991180Z ======================================================================
2018-12-31T21:45:49.9991601Z FAIL: test_asarray (numba.tests.test_np_functions.TestNPFunctions)
2018-12-31T21:45:49.9992124Z ----------------------------------------------------------------------
2018-12-31T21:45:49.9992424Z Traceback (most recent call last):
2018-12-31T21:45:49.9992675Z   File "/home/vsts/work/1/s/numba/tests/support.py", line 643, in tearDown
2018-12-31T21:45:49.9992890Z     self.memory_leak_teardown()
2018-12-31T21:45:49.9993125Z   File "/home/vsts/work/1/s/numba/tests/support.py", line 617, in memory_leak_teardown
2018-12-31T21:45:49.9993342Z     self.assert_no_memory_leak()
2018-12-31T21:45:49.9993713Z   File "/home/vsts/work/1/s/numba/tests/support.py", line 626, in assert_no_memory_leak
2018-12-31T21:45:49.9993929Z     self.assertEqual(total_alloc, total_free)
2018-12-31T21:45:49.9994116Z AssertionError: 3 != 2

[stuartarchibald]

@seibert Think it needs #3612, will merge that and then see if it is happier.

[seibert]

This seems to be passing Azure now, except for a random failure in the Windows Py27 environment caused by a failure to install VS2008.

[stuartarchibald]

Build seems to be ok now as noted :)

[seibert]

Looks good.