Preceding text is discarded if acronym follows a non-breaking space

[jschlatow]

Pandoc automatically inserts non-breaking spaces after abbreviations such as "e.g." and "i.e.".
As a consequence, the words processed by the Str function of this filter may actually contain whitespace characters.
However, the filter merely matches printable characters (not including spaces) via %g* to identify the text before/after the acronym.
As a result, text before a non-breaking space will be discarded.

This problem is reproducible by the following example:

Some text (e.g. (+FOOBAR))).

Which will produce the following LaTeX output:

Some text \gls{FOOBAR}.

I'd suggest to use a .* instead of a %g* to match the front and back part.