_diff_text: use repr with escape characters

[blueyed]

This helps especially with comparing strings containing terminal escape
sequences, but also newlines already.

Ref: #3443

TODO:

• 8de826c

[The-Compiler]

I'm a bit confused by this change (and others) - do you consider a newline a non-printable/zero-width character? Why? It seems quite confusing to me to see a multi-line output but also \n, i.e. with newlines represented twice.

[blueyed]

@The-Compiler

Why?

Otherwise it will not be visible. Check https://github.com/pytest-dev/pytest/pull/6702/files/9318b236ff35fc6a0734ee91bff2011afb93971a#diff-14ad89cac128f1dd3aaa927c656035b3R1342.

Note: this is for a line (text/str) diff.

[blueyed]

It could be made smarter though maybe to not show them when not at the end of a line?

[nicoddemus]

For illustration, on master:

    def test_eq_similar_text():
        x ="foo\n1 bar"
>       assert x == "foo\n2 bar"
E       AssertionError: assert 'foo\n1 bar' == 'foo\n2 bar'
E           foo
E         - 1 bar
E         ? ^
E         + 2 bar
E         ? ^

On this branch:

    def test_eq_similar_text():
        x ="foo\n1 bar"
>       assert x == "foo\n2 bar"
E       AssertionError: assert 'foo\n1 bar' == 'foo\n2 bar'
E         NOTE: Strings contain non-printable/zero-width characters. Escaping them using repr().
E           'foo\n'
E         - '1 bar'
E         ?  ^
E         + '2 bar'
E         ?  ^

Is this what you would prefer @The-Compiler ?

    def test_eq_similar_text():
        x ="foo\n1 bar"
>       assert x == "foo\n2 bar"
E       AssertionError: assert 'foo\n1 bar' == 'foo\n2 bar'
E         NOTE: Strings contain non-printable/zero-width characters. Escaping them using repr().
E         - 'foo\n1 bar'
E         ?       ^
E         + 'foo\n2 bar'
E         ?       ^

[blueyed]

For illustration, on master:

    def test_eq_similar_text():
        x ="foo\n1 bar"
>       assert x == "foo\n2 bar"
E       AssertionError: assert 'foo\n1 bar' == 'foo\n2 bar'
E           foo
E         - 1 bar
E         ? ^
E         + 2 bar
E         ? ^

On master it has a different order (which I find still wrong - it should be left-to-right, top-to-bottom, with only +/- swapped, but off topic here - maybe therefore edited manually?):

    def test_eq_similar_text(self):
        x = "foo\n1 bar"
>       assert x == "foo\n2 bar"
E       AssertionError: assert 'foo\n1 bar' == 'foo\n2 bar'
E           foo
E         - 2 bar
E         ? ^
E         + 1 bar
E         ? ^

How about?

    def test_eq_similar_text(self):
        x = "foo\n1 bar\n"
>       assert x == "foo\n2 bar"
E       AssertionError: assert 'foo\n1 bar\n' == 'foo\n2 bar'
E         NOTE: Strings contain different line-endings. Escaping them using repr().
E         - 'foo\n1 bar\n'
E         ?       ^    --
E         + 'foo\n2 bar'
E         ?       ^

However, with longer strings it is useful to split them on newlines, of course.

>       assert x == "foo\n1 bar"
E       AssertionError: assert 'foo\n1 bar\n' == 'foo\n1 bar'
E         NOTE: Strings contain different line-endings. Escaping them using repr().
E           foo
E         - 1 bar\n
E         ?      --
E         + 1 bar
E         -

FWIW it always looked a bit strange to me seeing:

- foo
+ foo
     ^

(it could also be a space etc)

E         - 'foo\n1 bar'
E         ?       ^
E         + 'foo\n2 bar'
E         ?       ^

This could also be triggered via some minimal length (related: blueyed#218, where I split it onto separate lines with a certain length).

[The-Compiler]

Hmm, I guess explicit is indeed better than implicit in this case. I agree having the ^ marker pointing to "nothing" is odd, and I remember people being confused about that.

[blueyed]

WIP to handle different line endings only