GraphViz::parse_string and unicode lables

[vizvamitra]

I've found some strange behavior:

dot_str = "digraph G {\nHello [label = \"안녕하세요\"];\nWorld [label = \"Мир\"];\n  Hello -> World;\n}\n"

GraphViz.parse_string(dot_str).output(none: String)
# => "digraph G {\nHello [label = \"\\xEC\\x95\\x88\\xEB\\x85\\x95\\xED\\x95\\x98\\xEC\\x84\\xB8\\xEC\\x9A\\x94\"];\nWorld [label = \"\\xD0\\x9C\\xD0\\xB8\\xD1\\x80\"];\n  Hello -> World;\n}\n""

GraphViz.parse_string(dot_str).output(png: 'test.png') # =>

(ruby 2.0.0p353, ruby-graphviz 1.2.2, graphviz 2.36.0)

[froderik]

Same issue here but with swedish characters. Any progress? As far as I can tell GraphViz supports unicode so I guess it is called in the wrong way somehow.

[vizvamitra]

@froderik I ended up writing my own simple implementation of a tool that converts directed graph to dot string and feeding this string directly to dot command.

[froderik]

aah - thanks for posting! We have a dot file already that is generated from another source so maybe we should try to just pass it directly to dot instead of using graphviz. What do you think?

[vizvamitra]

@froderik it does work with russian so I think it'll do so with swedish too =)

system('dot /path/to/dot/file.dot -Tsvg -o /path/to/output/file.svg')

But you will probably need to handle all posible errors from dot command (I believe in this gem they handle errors here with just raising "Error from #{cmd}:\n#{errors}")

Also Ruby-Graphviz gem is quite heavy and I guess it'll be beter not to use it if your use cases are as simple as just format simple graph to dot or convert dot to image

[froderik]

great! will try it right away!

[froderik]

@vizvamitra - works like a charm - thanks for the help!

[abruzzi]

Any update about this?

I got a similar issue with Chinese character in the label of nodes.

script as the following

require "graphviz"

g = GraphViz::new( "G" )
g.add_nodes "chinese", label: '你好'
g.output( :dot => "x.dot" )

would generate dot as this:

digraph G {
    graph [bb="0,0,115.69,36"];
    node [label="\N"];
    chinese  [height=0.5,
        label="\u4F60\u597D",
        pos="57.845,18",
        width=1.6068];
}

and will generate graph contains node with label as u4F60u597D, any solutions or walk around please ?

[atitan]

It's platform-related issue. to_gv uses 'string'.to_s.inspect.gsub( "\\\\", "\\" ) to escape string, and inspect behaves differently on Windows.

c:\script>ruby test.rb
"\u4F60\u597D"
[user@i2 ~]$ ruby test.rb
"你好"

reference: http://stackoverflow.com/questions/34454084/ruby-sometimes-prints-unicode-escaped-chars-instead-of-the-chars-themselves-why

[ghost]

The problem is in lib/graphviz/utils.rb:46 function output_and_errors_from_command: Open3 works in binmode, that force Encoding:ASCII-8BIT

[davidmoshal]

this worked for me: < for the less than sign.

[vanboom]

This issue is still present in 2021 when labels have UTF-8 character codes, in our case French names with é character.

[vanboom]

The problem is in lib/graphviz/utils.rb:46 function output_and_errors_from_command: Open3 works in binmode, that force Encoding:ASCII-8BIT

When producing a SVG using -Tsvg, binode should be false. Working on a PR with hopes to finally resolve this. Is this gem actively maintained?

[vanboom]

PR submitted - not sure if anyone is listening though.

To contributors: Thank you for your consideration on my PR and for supporting this great gem.