[Pyparsing] setResultsName on a recursive element of grammar

[Paul McGuire]

Elizabeth -

Googling for RFC1429, I found this BNF, which looks like what you are
working from:

https://tools.ietf.org/html/rfc1459#section-2.3.1


<message> ::= [':' <prefix> <SPACE> ] <command> <params> <crlf>
<prefix> ::= <servername> | <nick> [ '!' <user> ] [ '@' <host> ]

<command> ::= <letter> { <letter> } | <number> <number> <number>
<SPACE> ::= ' ' { ' ' }
<params> ::= <SPACE> [ ':' <trailing> | <middle> <params> ]

<middle> ::= <Any *non-empty* sequence of octets not including SPACE
or NUL or CR or LF, the first of which may not be ':'>
<trailing> ::= <Any, possibly *empty*, sequence of octets not including
NUL or CR or LF>

<crlf> ::= CR LF

From this BNF, I came up with this translation to pyparsing, very similar to

yours:

COLON = Suppress(':')
command = Word(alphas) | Word(nums, exact=3)
middle = ~COLON + Word(printables)
trailing = Word(printables)

params = Forward()
params <<= COLON + trailing | middle + params

I usually leave the assignment of results names until the very end, just
assigning them in the expressions where they get composed into groups or the
top-most parse expression.

line = command("command") + Group(params)("params")

tests = """\
COMMAND param1 param2 : param3"""
line.runTests(tests)

And this gives:

COMMAND param1 param2 : param3
['COMMAND', ['param1', 'param2', 'param3']]
- command: COMMAND
- params: ['param1', 'param2', 'param3']

This is something of a problem, since we have lost the distinction of which
part of the params are the middle and which are the trailing. The issue is
that recursive definition of params, which you pointed out makes the results
awkward to work with. The best I could do here was to define params using:

params <<= (COLON + trailing("trailing") |
middle("middle*") + params)

(I'm using the abbreviated version of `setResultsName`, using the
expressions as callables - the trailing '*' in "middle*" is equivalent to
`middle.setResultsName("middle", listAllMatches=True)`. And as you probably
already discovered, if `listAllMatches` is left out, then you will only get
the last element of `middle`.)

With this change, I get:

COMMAND param1 param2 : param3
['COMMAND', ['param1', 'param2', 'param3']]
- command: COMMAND
- params: ['param1', 'param2', 'param3']
- middle: [['param1'], ['param2']]
[0]:
['param1']
[1]:
['param2']
- trailing: param3

Which is *okay* but not really pleasant to deal with that middle bit.

But I'd like to look at this recursive construct in the original BNF:

<params> ::= <SPACE> [ ':' <trailing> | <middle> <params> ]

This is very typical in many BNFs, which will define a repetition of one or
more items as:

<list_of_items> ::= <item> [ <list_of_items> ]

This *can* be implemented in pyparsing as:

list_of_items = Forward()
list_of_items <<= item + list_of_items

But you'll find in pyparsing that things are usually clearer (and faster)
when you define repetition using the OneOrMore or ZeroOrMore classes:

list_of_items = OneOrMore(item)

If we use a repetition expression instead of a recursive expression for
params, it looks like this:

params = (OneOrMore(middle)("middle") +
COLON +
ZeroOrMore(trailing)("trailing"))

And the parsed test string gives:

COMMAND param1 param2 : param3
['COMMAND', ['param1', 'param2', 'param3']]
- command: COMMAND
- params: ['param1', 'param2', 'param3']
- middle: ['param1', 'param2']
- trailing: ['param3']

Here is the whole parser in one copy/pasteable chunk:

command = Word(alphas) | Word(nums)
COLON = Suppress(':')
middle = ~COLON + Word(printables)
trailing = Word(printables)
params = (OneOrMore(middle)("middle") +
COLON +
ZeroOrMore(trailing)("trailing"))
line = (command("command") + Group(params)("params"))

tests = """\
COMMAND param1 param2 : param3"""
line.runTests(tests)

And no need to kludge in any `listAllMatches` behavior either.

-- Paul



---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

[Elizabeth Myers]

Post by Paul McGuire
command = Word(alphas) | Word(nums)
COLON = Suppress(':')
middle = ~COLON + Word(printables)
trailing = Word(printables)
params = (OneOrMore(middle)("middle") +
COLON +
ZeroOrMore(trailing)("trailing"))
line = (command("command") + Group(params)("params"))
tests = """\
COMMAND param1 param2 : param3"""
line.runTests(tests)
And no need to kludge in any `listAllMatches` behavior either.
-- Paul
---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

There was one minor thing I forgot. Word(printables) is insufficient in
your example, as any 8-bit string is acceptable in parameters (and are
often used).

The encoding is not specified for these portions (deliberately it
seems), but I usually implicitly assume UTF-8, since that is the
de-facto standard. As such, UTF-8 is what I decode all stuff from the
wire to, using the 'replace' error handler.

This is my proposed solution to match all characters but surely there is
a better way than the below (perhaps using Regex is better?):

# 1111998 is the total number of valid Unicode characters.
utf8_chars = ''.join(chr(x) for x in range(1111998))
middle = (~COLON + ~White()) + Word(utf8_chars)

--
Elizabeth