Preserve Float Subclass

[deckar01]

Numbers are cast to float before format()ing, which prevents custom formatting logic provided by subclasses. tabulate/__init__.py#L1350

Example

from tabulate import tabulate
from prefixed import Float

tabulate([[Float(1)]], floatfmt='h')

ValueError: Unknown format code 'h' for object of type 'float'

Background

Proposals to extend the built-in python formats have been rejected, because subclasses are intended to satisfy this use case.

https://discuss.python.org/t/new-format-specifiers-for-string-formatting-of-floats-with-si-and-iec-prefixes/26914/6

This regression was introduced in version 0.2. 5115cea/tabulate.py#L154

Related to #297.

Proposal

Only apply the float cast in _format() to strings.

Edit: And decimal align any string by . or \d([^d]) in _afterpoint().

[deckar01]

After patching that I also found that the default decimal alignment does not behave as expected. _afterpoint() requires the string to pass _isnumber(), which can fail for custom formats. I would propose removing this restriction and allowing arbitrary strings to be aligned by decimal.

[deckar01]

I also had to generalize the e position logic in _afterpoint() to get custom suffixes aligned correctly without decimal points. Aligning on the rightmost digit of the string regardless of the next character aligns prefixes, suffixes, and sandwiches 🥪 (like e).

  f0     Q    gain    Av    BW    C1     R1    R2      R3
----  ----  ------  ----  ----  ----  -----  ----  ------
 700  2      10     3.16   350  100n  1.44k   940   9.09k
 701  2       9.64  3.03   350  100n  1.5k    910   9.1k
 728  2.52   13.2   4.58   289  100n  1.2k    680  11k
 719  2.71   12     4      265  100n  1.5k    560  12k

[deckar01]

master...deckar01:python-tabulate:361-custom-format

https://pypi.org/project/deckar01-tabulate/0.10.2/