string escapes can lead to parse errors via Python API

[abathur]

Just want to document a sharp corner I ran into while using this. Not sure there's a fix. (It looks like sedparse is parsing correctly and IDK if there's a reliable way to normalize the input in all cases.)

I'm handling a script that has something like the below (I'm simplifying it to just the expression that triggers the error):

sed -E -e "s,.*(query\\[A|DHCP).*,${COL_NC}&${COL_NC},"

I'm using argparse to identify the expressions and just pass those to sedparse, and when it tries to parse the expression above it raises:

...
  File "/nix/store/q0fjmh9bjn7shcbrmcnir34bxlbs7a81-resholve-0.9.1-dirty/bin/.resholve-wrapped", line 3726, in _find_sed_e_cmd
    sedparse.compile_string(parsed, expr)
  File "/nix/store/py54k9w06iqh5x2pssv87mwpmjh702g5-python2.7-sedparse-0.1.2/lib/python2.7/site-packages/sedparse.py", line 1408, in compile_string
    compile_program(cur_program)
  File "/nix/store/py54k9w06iqh5x2pssv87mwpmjh702g5-python2.7-sedparse-0.1.2/lib/python2.7/site-packages/sedparse.py", line 1321, in compile_program
    bad_prog(UNTERM_S_CMD)
  File "/nix/store/py54k9w06iqh5x2pssv87mwpmjh702g5-python2.7-sedparse-0.1.2/lib/python2.7/site-packages/sedparse.py", line 648, in bad_prog
    exitcode=EXIT_BAD_USAGE,
sedparse.ParseError: sedparse: -e expression #3, char 35: unterminated `s' command

I'm blaming escaping because of the behavior I see when I spot-check this from the CLI.

The parse error doesn't happen if I feed a similar string in to the CLI:

$ echo heh | sed -e "s,.*(query\\[A|DHCP).*,&,"
heh

$ ... /sedparse.py -e "s,.*(query\\[A|DHCP).*,&,"
[
    {
        "cmd": "s",
        "line": 1,
        "x": {
            "cmd_subst": {
                "regx": {
                    "pattern": ".*(query\\[A|DHCP).*",
                    "slash": ","
                },
                "replacement": {
                    "text": "&"
                }
            }
        }
    }
]

But if I switch to single quotes, sed and sedparse both emit the same parse error:

$ echo heh | sed -e 's,.*(query\\[A|DHCP).*,&,'
sed: -e expression #1, char 25: unterminated `s' command

$ ... /sedparse.py -e 's,.*(query\\[A|DHCP).*,&,'
sedparse: -e expression #1, char 25: unterminated `s' command

The single-quoted form works in both cases (and the parse tree is the same) if I remove one backslash:

$ echo heh | sed -e 's,.*(query\[A|DHCP).*,&,'
heh

$ ... /sedparse.py -e 's,.*(query\[A|DHCP).*,&,'
[
    {
        "cmd": "s",
        "line": 1,
        "x": {
            "cmd_subst": {
                "regx": {
                    "pattern": ".*(query\\[A|DHCP).*",
                    "slash": ","
                },
                "replacement": {
                    "text": "&"
                }
            }
        }
    }
]