Fix #5693: Implemented is_sorted in C

[mdboom]

No description provided.

[efiring]

Nice!

[tacaswell]

tad worried what this will do with large int arrays or pd.Series but so long an np will do the corrosion it should be no worse than it is now (I think).

[mdboom]

tad worried what this will do with large int arrays or pd.Series but so long an np will do the corrosion it should be no worse than it is now (I think).

Yeah. I think a conversion was already happening in the nanmin call (since nans don't even make sense with ints), so this is no worse. That said, there's probably some low-hanging fruit to do the easy thing if it's an int array to begin with.

[mdboom]

I've updated this to work on ints directly as well.

[tacaswell]

backport on this had conflicts, making sure it works locally before pushing

[tacaswell]

backported as db80928

[QuLogic]

I know this is merged already, but I'm curious as to the expected behaviour with respect to infinities? As you may guess from the name, std::isfinite returns false for infinity, so the behaviour is different, and there are no tests that exercise that case to show what's expected.

>>> import numpy as np
>>> x = np.arange(10.)
>>> x[4] = np.inf
>>> x
array([  0.,   1.,   2.,   3.,  inf,   5.,   6.,   7.,   8.,   9.])
>>> np.nanmin(x[1:] - x[:-1]) >= 0
False
>>> from matplotlib import _path
>>> _path.is_sorted(x)
True

Is it just not possible for there to be infinities at this location?

[tacaswell]

That is a good point that I think we missed on review.

On Wed, Dec 23, 2015 at 6:44 PM Elliott Sales de Andrade <
notifications@github.com> wrote:

I know this is merged already, but I'm curious as to the expected
behaviour with respect to infinities? As you may guess from the name,
std::isfinite returns false for infinity, so the behaviour is different,
and there are no tests that exercise that case to show what's expected.

import numpy as np>>> x = np.arange(10.)>>> x[4] = np.inf>>> x
array([ 0., 1., 2., 3., inf, 5., 6., 7., 8., 9.])>>> np.nanmin(x[1:] - x[:-1]) >= 0False>>> from matplotlib import _path>>> _path.is_sorted(x)True

Is it just not possible for there to be infinities at this location?

—
Reply to this email directly or view it on GitHub
#5719 (comment)
.