Fixing errors when parsing multiline outputs of xdotdata

[jimon]

On my Windows 10 computer with Graphviz 2.49.1 under ConEmu (not sure if this is relevant) translating a simple dot file via dot2texi:

\begin{dot2tex}[dot]
digraph G {
	rankdir=LR;

	closed [label = "Doors closed"];
	open [label = "Doors open"];

	closed -> open [label = "Paid"];
	open -> closed [label = "Passed"];
	closed -> closed [label = "Approached"];
}
\end{dot2tex}

Renders the following output from xdotdata:

2021-10-17 21:20:59,489 dot2tex DEBUG xdotdata:
b'digraph G {\r\n\tgraph [_draw_="c 9 -#fffffe00 C 7 -#ffffff P 4 0 0 0 71.7 303.48 71.7 303.48 0 ",\r\n\t\tbb="0,0,303.48,71.697",\r\n\t\trankdir=LR,\r\n\t\txdotversion=1.7\r\n\t];\r\n\tnode [label="\\N"];\r\n\tclosed\t[_draw_="c 7 -#000000 e 59.79 20.7 59.59 18 ",\r\n\t\t_ldraw_="F 14 11 -Times-Roman c 7 -#000000 T 59.79 17 0 76 12 -Doors closed ",\r\n\t\theight=0.5,\r\n\t\tlabel="Doors closed",\r\n\t\tpos="59.795,20.697",\r\n\t\twidth=1.661];\r\n\tclosed -> closed\t[_draw_="c 7 -#000000 B 7 42.66 38.11 39.46 47.79 45.17 56.7 59.79 56.7 68.94 56.7 74.59 53.22 76.77 48.24 ",\r\n\t\t_hdraw_="S 5 -solid c 7 -#000000 C 7 -#000000 P 3 80.27 48.16 76.93 38.11 73.28 48.05 ",\r\n\t\t_ldraw_="F 14 11 -Times-Roman c 7 -#000000 T 59.79 60.5 0 70 10 -Approached ",\r\n\t\tlabel=Approached,\r\n\t\tlp="59.795,64.197",\r\n\t\tpos="e,76.933,38.107 42.656,38.107 39.457,47.785 45.17,56.697 59.795,56.697 68.935,56.697 74.595,53.216 76.773,48.242"];\r\n\topen\t[_draw_="c 7 -#000000 e 249.53 20.7 53.89 18 ",\r\n\t\t_ldraw_="F 14 11 -Times-Roman c 7 -#000000 T 249.53 17 0 67 10 -Doors open ",\r\n\t\theight=0.5,\r\n\t\tlabel="Doors open",\r\n\t\tpos="249.53,20.697",\r\n\t\twidth=1.4985];\r\n\tclosed -> open\t[_draw_="c 7 -#000000 B 4 119.68 20.7 140.57 20.7 164.15 20.7 185.38 20.7 ",\r\n\t\t_hdraw_="S 5 -solid c 7 -#000000 C 7 -#000000 P 3 185.58 24.2 195.58 20.7 185.58 17.2 ",\r\n\t\t_ldraw_="F 14 11 -Times-Roman c 7 -#000000 T 157.59 24.5 0 25 4 -Paid ",\r\n\t\tlabel=Paid,\r\n\t\tlp="157.59,28.197",\r\n\t\tpos="e,195.58,20.697 119.68,20.697 140.57,20.697 164.15,20.697 185.38,20.697"];\r\n\topen -> closed\t[_draw_="c 7 -#000000 B 10 209.87 8.38 199.5 5.62 188.21 3.09 177.59 1.7 159.96 -0.62 155.24 -0.45 137.59 1.7 129.54 2.68 121.13 4.22 112.94 \\\r\n6 ",\r\n\t\t_hdraw_="S 5 -solid c 7 -#000000 C 7 -#000000 P 3 112.05 2.61 103.1 8.28 113.64 9.43 ",\r\n\t\t_ldraw_="F 14 11 -Times-Roman c 7 -#000000 T 157.59 5.5 0 40 6 -Passed ",\r\n\t\tlabel=Passed,\r\n\t\tlp="157.59,9.1973",\r\n\t\tpos="e,103.1,8.2833 209.87,8.3842 199.5,5.6207 188.21,3.0947 177.59,1.6973 159.96,-0.62201 155.24,-0.45332 137.59,1.6973 129.54,2.6778 \\\r\n121.13,4.2157 112.94,6.002"];\r\n}\r\n'

Which in result fails in dot2tex as following:

ERROR    Failed to process input
Traceback (most recent call last):
  File "C:\Python310\lib\site-packages\dot2tex\dot2tex.py", line 3075, in main
    s = conv.convert(dotdata)
  File "C:\Python310\lib\site-packages\dot2tex\dot2tex.py", line 832, in convert
    self.do_edges()
  File "C:\Python310\lib\site-packages\dot2tex\dot2tex.py", line 2266, in do_edges
    s += self.draw_edge(edge)
  File "C:\Python310\lib\site-packages\dot2tex\dot2tex.py", line 2295, in draw_edge
    pp.append("(%sbp,%sbp)" % (smart_float(p[0]), smart_float(p[1])))
  File "C:\Python310\lib\site-packages\dot2tex\dot2tex.py", line 169, in smart_float
    number_as_string = "%s" % float(number)
ValueError: could not convert string to float: '\\\r\n121.13'

My PR might be fixing just symptoms of the bigger issue with dot2tex parsing, but at least it unblocks me for now.
Let me know if it would be better to fix it in another place.

[GregoryMorse]

It looks like this is a symptom of the Popen not specifying text=True. These number parsing bugs are all over the place not just the one in question and removing the extra whitespace is not the ideal strategy. The stdout should be opened in text mode not binary mode as the code expects. Therefore this PR should be changed based on this though Python 2/3 versioning etc needs to be checked to see when the parameter was added and how its default worked, etc. encoding was added in Python 3.6 and text in Python 3.7.