/* *************************************************************************************
* copyright: Copyright (c) 2021 Lowell D. Thomas, all rights reserved
* license: BSD-2-Clause (https://opensource.org/licenses/BSD-2-Clause)
* ********************************************************************************* */This is a demonstration of the caveat about using rules and UDTs in look behind phrases.
The caveat: when in look-behind mode, the parser is actually parsing right to left instead of the normal left to right.
For all operators except rules (RNM) and UDTs (UDT) this is handled automatically by the parser.
But rules and UDTs of any complexity are likely to be dependent on the left to right direction in order to work correctly.
The example here is to match the character $ at string position 14, but only if it is preceded by a comment.
This example looks at using a comment rule that is left-to-right dependent and how it fails in look-behind.
It then looks at how the problem could be fixed by writing a special right-to-left version
of the comment rule especially for use in look-behind mode.
Again, this is a very artificial example here,
but is simulates a very real kind of situation that could arise
in the context of a pattern matching engine (see apg-exp).
What to expect:
Note that in the case of UDTs, there is a flag in the system data which indicates whether the parser is in look-behind mode or not, but in general you may not want to assume that that flag has been honored or maybe more importantly whether it has been honored correctly.
(function comment() {
let input = ' not a comment$';
const grammar = new (require('./comment-grammar'))();
const setup = require('./setup');
setup(input, grammar, 'comment-fail');
input = ' /* comment */$';
setup(input, grammar, 'comment-success');
})();