Frank Dana

> See point 1 in: https://github.com/pyparsing/pyparsing/wiki/Performance-Tips > > It's an easily toggleable feature. My quick tests in [#221](https://github.com/pydot/pydot/issues/221) showed that it speeds things up by at least 30% in a...

> You're justified in being surprised, I think I was measuring a PNG save by accident... > Of course a function that parses things spends ~99.99% of the time parsing....

> Maybe what we should add to the API are a set of static functions which can accept `dot`-syntax data (as a file on disk, or directly in a string),...

Heh. And I just realized: This would allow passing the result of `graph_from_dot_data()` directly to the renderer, even though it's a list. _Except_, the **whole point** of the static renderers...

\*nod\* I was thinking `filenames` could either take a string or a list of strings (or generator that produces strings, etc.); if it's a single string and there are multiple...

For now I've worked around this by setting the `TOX_GH_MAJOR_MINOR` environment variable to `3.13t` in the freethreaded instance, which works fine and causes it to match the corresponding mapping in...

(To save anyone looking to support this some time, recent versions of the `virtualenv.discovery.py_info.PythonInfo` class that tox-gh uses to query parameters of the various Python interpreters now include a `free_threaded`...

Yeah, the entire parser is implemented in the fashion of graphviz's own, which is notoriously not thread-safe and practically _made_ of global variables. Personally, I'd prefer to see the parser...

(It'd also make debugging f---loads easier if we could actually access all of the components of the parser, from outside the module. Right now, everything except the global `graphparser` is...

> Personally, I'd prefer to see the parser in `dot_parser.py` turned into a class, with each parsing run performed by a new instance of that class, all of the functions...