Import Markup and escape directly from MarkupsSafe

[rwedge]

Fixes #1736

Jinja 3.1.0 removed jinja.Markup and jinja.escape and suggested importing them from MarkupSafe

[rwedge]

Test failure appears related to Jinja2 3.1.0 release. Pin Jinja2 below 3.1.0 and the tests pass.

Possibly related to this change? pallets/jinja#1621

[trungleduc]

should we bump the lower bound and add an upper bound for jinja2 version?

[rwedge]

I set an upper bound for jinja2. Perhaps a different PR could investigate the windows errors?

[effigies]

This also closes #1605.

[rwedge]

@davidism could there have been a change in jinja2 behavior between 3.0 and 3.1 that would account for the new failures in these tests?

    def test_absolute_template_dir(self):
        with tempdir.TemporaryDirectory() as td:
            template = 'mytemplate'
            template_file = os.path.join(td, template, 'index.py.j2')
            template_dir = os.path.dirname(template_file)
            os.mkdir(template_dir)
            test_output = 'absolute!'
            with open(template_file, 'w') as f:
                f.write(test_output)
            config = Config()
            config.TemplateExporter.template_name = template
            config.TemplateExporter.extra_template_basedirs = [td]
            exporter = self._make_exporter(config=config)
>           assert exporter.template.filename == template_file
E           AssertionError: assert 'C:\\Users\\R...e/index.py.j2' == 'C:\\Users\\R...\\index.py.j2'
E             - C:\Users\RUNNER~1\AppData\Local\Temp\tmpd72mcthi\mytemplate\index.py.j2
E             ?                                                            ^
E             + C:\Users\RUNNER~1\AppData\Local\Temp\tmpd72mcthi\mytemplate/index.py.j2
E             ?     

https://github.com/jupyter/nbconvert/runs/5689650683?check_suite_focus=true

[davidism]

We did fix some issues with unsafe Windows paths in template names. The two filenames being compared are in fact equivalent on Windows, even though they're not textually equivalent, since / is a valid path separator. A quick workaround for you would be to use pathlib.Path() or os.path.normpath() on both values, which will normalize separators, which is what I'll do in a bugfix for Jinja as well.

[davidism]

Created pallets/jinja#1637, I will submit a PR and then release 3.1.1.

[davidism]

Released Jinja2 3.1.1 with the filename fix.

[SylvainCorlay]

Thanks everyone.

[SylvainCorlay]

I wonder if we should exclude 3.1.0 (while accepting 3.1.1) from the versions in setup.py.