lxml:xml_int_float_parsing

Last commit made on 2021-08-12
Get this branch:
git clone -b xml_int_float_parsing https://git.launchpad.net/lxml

Branch merges

Branch information

Name:
xml_int_float_parsing
Repository:
lp:lxml

Recent commits

83e6c03... by scoder

Implement a dedicated int/float parser for XML (schema) values in lxml.objectify.
This disables support for "_" in numbers, which are allowed by Python but not by XMLSchema.
Wee keep a few additional literals, such as "+NaN", simply because they shouldn't hurt.

See https://<email address hidden>/thread/6F7VIDKWZTJ6LB6VOX6IJNNWICYHFPNR/

e23a807... by scoder

Use Cython's autowrapping feature for cdef functions to keep internal utility functions out of the objectify module dict.

d866aad... by scoder

Remove outdated mention of Pyrex.

36bca0b... by scoder

Add note on crypto currency donations (and why we don't take them).

9f89e0f... by scoder

Update changelog.

8244dfd... by Petr Viktorin

_tofilelikeC14N: Always close output buffer (GH-322)

If `with writer.error_log` raises an exception, `c_buffer` would leak.
It seems that currently, it can't actually raise (it's uses small and tight `cdef` functions), but there's no guarantee they'll remain exception-free in the future.

But there's one more thing that potentially could leak (at least Cython generates an `unlikely` `goto` block for it):
the lookup of `__exit__` that happens at the start of the `with` block.

Put the `xmlOutputBufferClose` call into a `finally` block to make this safer.

02a49b1... by scoder

Rewrite Unicode chunk parsing by directly encoding to UTF-8.
Previously, we required Py_UNICODE strings, which is inefficient since most strings in Py3 use the PEP-393 memory layout.

b626841... by scoder

Try to get the wheel upload working in CI.

566effd... by scoder

Try to get the wheel upload working in CI.

7f03ec2... by scoder

Fix wheel build CFLAGS in CI.