pyparsing 3.0.2 broke pydot

[MarcCote]

Hi,

I just stumbled on this issue with pyparsing 3.0.2 (I can confirm 2.4.7, 3.0.0, and 3.0.1 were fine).

In [2]: import pydot

In [3]: pydot.graph_from_dot_data("""
   ...: graph G {
   ...:     b1 [label=<
   ...:          We are the knights who say <BR/># Ni
   ...:     >];
   ...: }
   ...: """)
graph G {
      ^
Expected '{', found 'G'  (at char 7), (line:2, col:7)

Note the leading whitespaces in the string.
Here's a simpler minimal example:

import pydot
pydot.graph_from_dot_data("graph G { b1 [label=<Hello World>]; }")

graph G { b1 [label=<Hello World>]; }
      ^
Expected '{', found 'G'  (at char 6), (line:1, col:7)

[ptmcg]

I'm looking at this now - can you post the expected output for both of these, so that if I resolve one part of the issue, I'll know if there is more to do, or the expected results were returned?

Thanks for checking this out so quickly!

[MarcCote]

Might not be super useful but here are the expected outputs:

In [1]: import pydot

In [2]: pydot.graph_from_dot_data("graph G { b1 [label=<Hello World>]; }")
Out[2]: [<pydot.Dot at 0x7f8c19cf8430>]

In [3]: pydot.graph_from_dot_data("""
   ...: graph G {
   ...:     b1 [label=<
   ...:          We are the knights who say <BR/># Ni
   ...:     >];
   ...: }
   ...: """)
Out[3]: [<pydot.Dot at 0x7f8c19d0fe80>]

pydot has some dependencies on graphviz, I've installed it through conda.
conda install graphviz

I'm running this version of graphviz.

$ dot -V
dot - graphviz version 2.40.1 (20161225.0304)

[ptmcg]

Yes, the pydot.Dot reprs aren't helping me much.

Testing the parser in isolation (with some changes in pyparsing), I get these outputs:

[['graph'], 'G', ['b1', 'label', '=', '<\n         We are the knights who say <BR/>\n    >']]
[['graph'], 'G', ['b1', 'label', '=', '<Hello World>']]

I can guess that the second one looks right, but the first one omits "# Ni" - is that the expected behavior?

Unfortunately, my pyparsing changes perturb some other behavior, so I'm not totally out of the woods. But it looks like I'm in the right place.

[ptmcg]

It looks like I have to undo a small performance optimization, but otherwise, all regression tests pass (plus the new ones added from this week to reference parsers that use pyparsing).

[MarcCote]

@ptmcg you are right about the "# Ni" it is expected at the moment (see open issue pydot/pydot#235). I shouldn't have used that example, my bad.

[ptmcg]

Fixed in 3.0.3, just pushed to pypi.

[MarcCote]

I can confirm 3.0.3 is working great for my use cases. Feel free to close this issue.