Custom grammar

[Ghesselink]

As an addition to the default behave grammar parsing. For example, to specify which values to permit
Given all instances of IfcGeometricRepresentationContext[without subtypes]

A solution would then be to use pyparsing

ifc_entity = Ifc\w+
grammar = ifc_entity + [ "with subtypes" | "without subtypes" ]

Advantages

• It's more descriptive in terms of what values it can accept
• It's stricter in making sure you don't supply wrong pattern
• You get much better error messages when it doesn't comply

The issue is also to link relevant code snippets/PR comments

[Ghesselink]

Applied in 6a125d3

[JTurbakiewicz]

Hi,

I am wondering, what are the advantages of using pyparsing vs using regex in the decorator, as in the below example:

https://behave.readthedocs.io/en/stable/api.html#step-parameters
https://jenisys.github.io/behave.example/step_matcher/re_matcher.html

The first thing that comes to my mind is that regex gets unreadable really quickly. With pyparsing this is perhaps less of an issue.

[Ghesselink]

I think you can reach the same with using regex as well, but pyparsing seems way more readable like you mention. There are some additional advantages that are probably/possibly also doable with regular expressions, but I found them quite intuitive with pyparsing.

e.g., it's possible to modify a parsed value immediately. For instance, the following line of code converts a string to an integer:
integer = Word(nums)("integer").setParseAction(lambda t:int(t[0])) One benefit of this is that the parsed sentence is saved in a dictionary, and the integer can be retrieved simply by calling parse['integer'].

It is also possible to group/nest statements. For example, in the following two sentences:

• Then Each "IfcAlignment" must be nested by exactly 1 instance(s) of "IfcAlignmentHorizontal"
• Then Each "IfcAlignmentHorizontal" is nested by a list of only instance(s) of "IfcAlignmentSegment"

Can be parsed with

condition_stmt = Optional(oneOf(conditions)('condition') | 
                        Group(Group(Word(alphas) + Optional(Word(alphas)))('operator') + integer('int'))('operator_stmt')
                        )

Then Each "IfcAlignment" must be nested by exactly 1 instance(s) of "IfcAlignmentHorizontal"
The values that have been parsed can be utilized in the Python implementation.

parse.get('operator_stmt', False) # exactly 1
stmt_to_op = {'exactly': operator.eq, "at most": operator.le}
constraint = ' '.join(operator_stmt['operator']) # exactly
num = operator_stmt['int'] # 1

The same for
Then Each "IfcAlignmentHorizontal" is nested by a list of only instance(s) of "IfcAlignmentSegment"
condition = parse.get('condition', False) # 'a list of only

There is also another example for the relationship_type (e.g. is nested by, nests, assigning, etc ..) in this PR #48. More specifically, in

ifc-gherkin-rules/features/steps/steps.py

Lines 643 to 646 in f4e5bc9

	relationship_type = Group(
	Or(map(Literal, list(conjugate_verbs(relationship_verbs))))('relationship_type') +
	Optional("by") #e.g. useful for the difference between 'Nests' and 'nestedBy'
	)('relationship_type')

.

Pyparsing also has support for nested statements like this by using recursion. I feel this would be a bit more robust and readable as there will be less 'Optional' statements/ and less dependency on order. But yeah, I am still trying that out/experimenting with it.

Something like :

relationship_fragment = Forward() # create forward reference for recursion
relationship_fragment << (condition_sentence | relationship_type | ) # | other type of grammar | other type of grammar

# define grammar for entire string
grammar = OneOrMore(relationship_fragment)

# parse the string using the grammar
parse = grammar.parseString(input_string)

[Ghesselink]

When considering the parsing of IFC versions and MVDs, is it possible to achieve this without using the " symbols, in order to maintain consistency?

ifc-gherkin-rules/features/SPS001_Basic-spatial-structure-for-buildings.feature

Lines 15 to 16 in 05329ef

	Given A file with Model View Definition "CoordinationView"
	And A file with Schema Identifier "IFC2X3"

[evandroAlfieri]

@aothms @Ghesselink is this still valid or obsolete?