On Sunday, 2 August 2015 at 18:22:01 UTC, Jacob Carlborg wrote:
On 02/08/15 19:15, Xinok wrote:
I guess you're not familiar with the theoretical aspect of
"formal
languages". The D grammar is a context-free grammar which
cannot be
reduced to a regular expression. As cym13 stated, there are
some simple
context-free grammars which can be rewritten as regular
expressions, but
the D grammar cannot be. Take a look at the Chomsky Hierarchy
[1] for a
better understanding.
The classic example of a context-free language is the set of
balanced
parenthesis, i.e. (()) is balanced and ())))) is not. This
language is
not regular meaning you cannot write a regular expression for
it, but
you can write a context-free grammar for it.
TextMate grammars are not _just_ regular expressions. They can
define balanced parentheses [1].
The point of a language grammar in a text editor is not to have
a 100% correct implementation of the grammar. Rather it should
syntax highlight the code in a way that is useful for the user.
[1] https://manual.macromates.com/en/language_grammars
Then your best shot is to approximate the grammar with the regual
expressions you have access to. You'll get to a point where some
constructs can not be correctly represented; at that point you
should probably write a regex which produces what the grammar
produces and some more.
In the example before of generating paired interleaved
parentheses, you could generate every possible combination of
parentheses, like
( (|)|[|]|{|}|" )*
where only the external parentheses are syntax for the regex.
That regex matches all the productions of the paired parentheses
grammar, and many more strings.
At the end of the day you want to highlight correct syntax, and
if an user writes wrong syntax is OK to have wrong highlight, so
be sure your regex work for the right syntax, and can do random
stuff for the wrong one
