Parse and group multiple items together using Pypa

Question

This is a build up on Build a simple parser that is able to parse different date formats using PyParse

I have a parser that should group one or more users together into a list So a.parser('show abc, xyz commits from "Jan 10,2015" to "27/1/2015"') should group the two usernames into a list [abc,xyz]

For users I have:

keywords = ["select", "show", "team", "from", "to", "commits", "and", "or"]
    [select, show, team, _from, _to,  commits, _and, _or] = [ CaselessKeyword(word) for word in keywords ]

user = Word(alphas+"."+alphas)
user2 = Combine(user + "'s")
users = OneOrMore((user|user2))

And the grammar is

bnf = (show|select)+Group(users).setResultsName("users")+Optional(team)+(commits).setResultsName("stats")\
    +Optional(_from + quotedString.setParseAction(removeQuotes)('from') +\
    _to + quotedString.setParseAction(removeQuotes)('to'))

This is erroneous. Can anyone guide me in the right direction. Also, is there a way in pyparse to selectively decide which group the word should fall under. What I mean is that 'xyz' standalone should go under my user list. But 'xyz team' should go under a team list. If the optional keyword team is provided then pyparse should group it differently.

I haven't been able to find what I am looking for online. Or maybe I haven't been framing my question correctly on Google?

Answer 1

You are on the right track, see the embedded comments in this update to your parser:

from pyparsing import *

keywords = ["select", "show", "team", "from", "to", "commits", "and", "or"]
[select, show, team, _from, _to,  commits, _and, _or] = [ CaselessKeyword(word) for word in keywords ]

# define an expression to prevent matching keywords as user names - used below in users expression
keyword = MatchFirst(map(CaselessKeyword, keywords))

user = Word(alphas+"."+alphas)  # ??? what are you trying to define here?
user2 = Combine(user + "'s")
# must not confuse keywords like commit with usernames - and use ungroup to 
# unpack single-element token lists
users = ungroup(~keyword + (user|user2))

#~ bnf = (show|select)+Group(users).setResultsName("users")+Optional(team)+(commits).setResultsName("stats") \
    #~ + Optional(_from + quotedString.setParseAction(removeQuotes)('from') +
                    #~ _to + quotedString.setParseAction(removeQuotes)('to'))

def convertToDatetime(tokens):
    # change this code to do your additional parsing/conversion to a Python datetime
    return tokens[0] 
timestamp = quotedString.setParseAction(removeQuotes, convertToDatetime)

# similar to your expression
# - use delimitedList instead of OneOrMore to handle comma-separated list of items
# - add distinction of "xxx team" vs "xxx"
# - dropped expr.setResultsName("name") in favor of short notation expr("name")
# - results names with trailing '*' will accumulate like elements into a single
#   named result (short notation for setResultsName(name, listAllValues=True) )
# - dropped setResultsName("stats") on keyword "commits", no point to this, commits must always be present
#
bnf = ((show|select)("command") + delimitedList(users("team*") + team | users("user*")) + commits + 
            Optional(_from + timestamp('from') + _to + timestamp('to')))

test = 'show abc, def team, xyz commits from "Jan 10,2015" to "27/1/2015"'

print bnf.parseString(test).dump()

Prints:

['show', 'abc', 'def', 'team', 'xyz', 'commits', 'from', 'Jan 10,2015', 'to', '27/1/2015']
- command: show
- from: Jan 10,2015
- team: ['def']
- to: 27/1/2015
- user: ['abc', 'xyz']