Modified insertions and joins for empty matrix #10815
Conversation
Previously it was converted to 1 due to logic that was guarding against None from the slice.
Fixes sympy#10606. The logic in Range._intersect for intersecting with intervals is particularly nasty, because of the corner cases with infinite ranges (ranges with infinite starts are handled differently). I believe the test cases test all the proper combinations, but the logic should still probably be checked, and it likely can be simplified. This changes the behavior of Range.__iter__ to raise an exception for ranges with infinite starts, since ranges always iterate in order. See the discussion on the issue for more information about this. I felt that it was better to always iterate in order, and catch possible bugs if the range is reversed, than to iterate backwards in this case. It's worth noting, however, that this makes it more annoying to determine the values in an infinite range beyond the two that are printed by the pretty printer. But to fix this, I think we should add better support for slicing ranges. Also added a better docstring to Range.
It was doing the wrong thing for infinite and reversed ranges.
This commit addresses the issue sympy#10770.
|
This should reference #10781. A better design issue there is that the same type of matrix is returned as was represented by the empty matrix at the start. |
|
Sir, I didn't get you. Can you please elaborate? |
| @@ -2082,6 +2084,9 @@ def test_row_insert(): | |||
| l = [1, 0, 0] | |||
| l.insert(i, 4) | |||
| assert flatten(eye(3).row_insert(i, r4).col(0).tolist()) == l | |||
| a = Matrix([1, 2]) | |||
| assert Matrix([]).row_insert(10, a) == a | |||
There was a problem hiding this comment.
I think for safety reasons, this should raise an error. If someone thinks they are working with a matrix that has size and they aren't, the error should not pass silently. Though a case could be made that list slices that contain out of range bounds are silently truncated as in L = [0, 1, 2]; L[:10] = [42] -> L == [42]. Do you see any merit in either approach?
|
Sir, why is it saying checks have failed. I have checked on my system before committing. Then if there are any errors, I think it would say DO NOT COMMIT. But it didn't say anything. So I have committed. What can I do now Sir? |
|
The checks have failed because the support for Python versions 2.6 and 3.2 was dropped a few days ago. |
|
Thank you, @jksuom . Have done the same 😄 |
| @@ -3983,6 +3983,8 @@ def row_join(self, rhs): | |||
| row | |||
| col_join | |||
| """ | |||
| if not self: | |||
| return rhs | |||
There was a problem hiding this comment.
here and below, this should be return self.func(rhs) (or similar) so the same type of matrix represented by self is returned. But I think the other duplicate PR will be doing this, so let's see.
update AUTHORS and .mailmap
|
@smichr Then what I have found is that I also have to change Thanks for the suggestion 👍 😄 |
Removed trailing whitespaces.
Direct test for is_deriv_k
Intersection.reduce short circuits on EmptySet() so Range(oo).intersect(S.EmptySet) -> S.EmptySet and happens outside of the Range methods. So rather than sometimes returning Range(0) and sometimes EmptySet, all Range intersections that are empty are returned as EmptySet. Slicing of the Range, however, can return Range(0) (which, unlike xrange/range, does not retain any information about how the empty Range was obtained). Range/Range intersections are now defined. Note about infinite values. --------------------------- Although selection of an infinite value in a Range is allowed, e.g. Range(oo)[-1] -> oo, slicing done from the end containing the infinite value is disallowed since 1) there is no way to count the nth value from oo, e.g. Range(oo)[-2:] 2) there is no way to represent a single infinite value while retaining normal Range semantics, Range(oo, oo + 1, 1) -> Range(oo, oo, 1) which would be the empty set. So although Range(oo)[-1:] could be represented as FiniteSet(oo) or an infinite step could be used and defined to give only the first value in a Range, e.g. Range(3,oo,oo) -> Range(3, 4, 1) hence list(Range(oo, 0, -oo)) -> [oo], neither of these seems like a good idea. As oo in S.Integers is False, oo in Range(oo) is also False; there is a slight iconsistency in letting Range(oo)[-1] give oo even though oo is not in the Range, but I don't see a reason to complicate the handling of inspection of the Range start/stop values.
Conflicts: sympy/printing/pretty/tests/test_pretty.py
Conflicts: sympy/printing/latex.py sympy/printing/pretty/pretty.py sympy/sets/fancysets.py sympy/sets/tests/test_fancysets.py
Added algorithm description
Reversed Range and Range.__getitem__
Modified algorithm description
Updated algorithm description
Improved spacing.
Fixed a typo
Fixed issue sympy#7021 (Faster primepi function)
added testcases
Updated primepi
|
Sir, I don't know what is happening. I am creating a new PR for only the modification of |
This solves the issue 10770. I think this is the best and simplest way to solve this. This just returns the argument matrix which we want to insert or join irrespective of the position which is the desired thing.