Formally specify PICA Path expression syntax

[nichtich]

PICA::Path should include a formal grammar of PICA Path expression syntax. The current implementation is:

    ([012.][0-9.][0-9.][A-Z@.])     # tag
    (\[([0-9.]{2,3}|\d+-\d+)\])?    # occurrence
    (\$?([_A-Za-z0-9]+))?           # subfields
    (\/(\d+)?(-(\d+)?)?)?           # position

And some additionals checks on occurrence and position ranges.

The syntax will be extended by conditions/subspects (#108) and x-counters (#97).

[nichtich]

Based on additions and different design choices by @niko2342 in pica-rs I tried to create a common subset to unify PICA Path expression syntax. The dev branch contains a new PICA Path expression syntax that should be like this:

path             ::=  tag occurrence? subfields? position?
tag              ::=  [012.] [0-9.] [0-9.] [A-Z@.]
occurrence       ::=  "/" occurenceValue | "[" occurrenceValue "]"
occurenceValue   ::=  number "-" number | occurrencePattern | "*"
occurencePattern ::=  [0-9.] [0-9.]? [0-9.]?
subfields        ::=  [$.]? ( [A-Za-z0-9]+ | "*" )
position         ::=  "/" ( number | range ) 
range            ::=  number "-" number? | "-" number
number           ::=  [0-9]+

Some features are not supported by pica-rs yet (deutsche-nationalbibliothek/pica-rs#346) and I am not sure whether occurrence in "[...]" and positions should be part of the core syntax or optional extensions, such as making the $ or . optional before subfield codes.

The introduction of / to start occurrences is a breaking change for pathes with position but no occurrence or subfields. To reference a position across all subfields add $* to explicitly refer to all subfields (e.g. 123X$*/3-7 or 123X*/3-7 instead of 123X/3-7).

In a path without occurrence, the occurrence is set to default value zero

• zero if first character of tag is 0 or 1 (because occurrence matters as field identifier on level 0 and 1)
• * (or 0-999) otherwise (to match all fields with or without occurrence)

This is also a breaking change. Until version 1.34 the path expression 0123 did also match 0123/01-99.

[nichtich]

See http://format.gbv.de/query/picapath for current state.