Fix false friends in implicit string concatenation in tests

[jbdyn]

In a PR about how to syntax-highlight Python docstrings, @rmuir and I discovered that instead of implicit concatenation of strings a string plus a docstring have been written.

Since this is a subtle one, I want to briefly show the differences:

The false friend of implicit string concatenation

# var == "foo"

var = "foo"  # <-- no implicit string concatenation
"bar"        # <-- docstring, legal for the bytecode compiler, against PEP 257

can be fixed for example with surrounding brackets:

# var == "foobar"

var = (
    "foo"    # <-- gets implicitly concatenated
    "bar"
)

I felt free to fix the ones I found right away such that the tests pass.

I searched with ripgrep like so:

# in path/to/cloned/pandas
rg -A 1 -B 2 -U ' = f?"[^"]*"\s+f?"[^"]+"\s*' pandas/

I am pretty confident, but not entirely sure, to have catched all cases. 🤔

[jbdyn]

As a side note:

The tests passed before because they called assert_produces_warning, which in turn calls _assert_caught_expected_warnings, where re.search is used:

pandas/pandas/_testing/_warnings.py

Lines 181 to 183 in 04356be

	if match is not None:
	if re.search(match, str(actual_warning.message)):
	matched_message = True

The false friends only gave the first part of the full string to match (the rest was discarded as it was detected as docstring), but re.search still gives a positive match in that case.

Maybe it should be re.fullmatch instead? 🤷

[rmuir]

I am pretty confident, but not entirely sure, to have catched all cases. 🤔

I rechecked the logfile from the script I ran yesterday: these are the same files and lines with highlight differences, so I think you found them all.

[mroeschke]

Thanks @jbdyn