=?UTF-8?Q?APG_=e2=80=93_ABNF_Parser_Generator=2c_Version_7=2e0?=

Lowell Thomas <ldt@sabnf.com>
Sun, 21 Feb 2021 16:35:08 -0500

          From comp.compilers

Related articles
=?UTF-8?Q?APG_=e2=80=93_ABNF_Parser_Generator=2c_Version_7=2e0?= ldt@sabnf.com (Lowell Thomas) (2021-02-21)
| List of all articles for this month |

From: Lowell Thomas <ldt@sabnf.com>
Newsgroups: comp.compilers
Date: Sun, 21 Feb 2021 16:35:08 -0500
Organization: Compilers Central
Injection-Info: gal.iecc.com; posting-host="news.iecc.com:2001:470:1f07:1126:0:676f:7373:6970"; logging-data="52791"; mail-complaints-to="abuse@iecc.com"
Keywords: parse, available
Posted-Date: 21 Feb 2021 17:05:32 EST

APG – ABNF Parser Generator, Version 7.0 is now available. APG generates
recursive-descent parsers directly from ABNF grammars and is therefore
well-suited to applications for Internet specifications which are often
defined with ABNF syntax. Translations are done via callback functions
in real time from the parse tree nodes or, optionally, at a later stage
from the AST nodes if you choose to generate one.


The source code is here (https://github.com/ldthomas/apg-7.0) and the
documentation is here (https://sabnf.com/documentation-2/). It is
licensed with the permissive 2-Clause BSD license so you can use it
pretty much as you like.


I’ve been away from this for a while and it is hard to believe that it
has been 16 years since the first version
(2005-06-04)(https://compilers.iecc.com/comparch/article/05-06-027) and
9 years since the last C version
(2012-07-02)(https://compilers.iecc.com/comparch/article/12-07-003). But
I wanted to bring it up to date and add a few features I’ve been
planning for a while. Actually, quite a few but I’ll just mention two or
three of the main new additions here.


Optionally, APG can generate and use Partially-Predictive Parsing Tables
(PPPTs). That is, from an examination of the ABNF grammar the generator
can determine the range of alphabet characters and generate a table, one
entry for each character and parse tree node. A PPPT entry can have one
of four values:
      • match – the node accepts the single character as a complete
phrase match
      • empty – the node does not accept the character but does accept an
empty string match
      • no match – the node rejects the character
      • active – the node accepts the character but not as a full phrase
match, parsing must continue normally
The entries are not just for the terminal nodes. The generator has rules
for walking back up the parse tree and generating a table entry for
every node, terminal and non-terminal alike. I some cases even the root
node can accept or reject a character without ever having to descend the
parse tree at all. As a general rule, I’ve found that PPPTs will
increase parsing speeds by a factor of 2.


APG is developed as an API so you can build custom generators and
generate parsers on the fly in your own applications.


It includes a Pattern-Matching Engine which I believe is more powerful
than regex.
      • replaces cryptic regex syntax with ABNF
      • full recursion can match deeply nested pairs
      • has two modes of back referencing. Introduces what I term
“parent-mode” back referencing. In particular it facilitates matching
not only the start and end tags of HTML or XML, with parent-mode back
referencing it is possible to match the tag names as well. I’m not
really a scholar on the topic so I won’t go so far as to say this has
never been done before, but I’m not aware of this type of back
referencing in any flavors of regex.
      • allows handwritten code snippets for difficult-to-define phrases
      • exposes the parser’s AST for complex translations of the matched
phrases
      • exposes a tracing facility which make debugging new pattern
syntax easy


There’s lots more, but if you are interested you can read about it in
the documentation.


Regards,
Lowell Thomas


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.