Is it the job of a parser to validate the input data?

Christopher F Clark <>
Thu, 12 Aug 2021 15:10:40 +0300

          From comp.compilers

Related articles
Is it the job of a parser to validate the input data? (Roger L Costello) (2021-08-11)
Is it the job of a parser to validate the input data? (Christopher F Clark) (2021-08-12)
Re: Is it the job of a parser to validate the input data? (George Neuner) (2021-08-12)
Re: Is it the job of a parser to validate the input data? (luser droog) (2021-09-03)
| List of all articles for this month |

From: Christopher F Clark <>
Newsgroups: comp.compilers
Date: Thu, 12 Aug 2021 15:10:40 +0300
Organization: Compilers Central
Injection-Info:; posting-host=""; logging-data="55169"; mail-complaints-to=""
Keywords: parse, syntax
Posted-Date: 12 Aug 2021 11:52:12 EDT

Roger L Costello <> asked:

> There are many data formats which contain things like this:
> A number, N
> N occurrences of something
> For example, 3 followed by the names of three students:
> 3
> John Doe
> Sally Smith
> Judy Jones
> I have a question about parsing such data. Is it the job of a parser to ensure
> that the number of student names matches the number? Or, is it the job of the
> parser to merely tokenize whatever is in the input and then create an abstract
> syntax tree containing the tokens?

It is almost always done in the AST creation routines, not only do you
as our insightful moderator mentioned generally get better error
messages that way, but curiously, the features of extract a number,
turn it into a count, and apply that count (and yes those might be 3
distinct operations) to be how many items a list involves has not been
implemented in any parser generator or lexer generator that I have
ever seen. That's a bizarre omission, particularly since it is a
common feature in many languages like networking protocols. Doing
fixed counts isn't rare, but doing a count held in a "register" or
"variable" seems to not be done.

The conversion step should generally be deferred to "semantic (aka
action) code or a predicate" as the process is messy and best handled
by some well tuned code not something a lexer/parser generator just
outputs and hopes it is semantically correct.

I have i(all 3 steps) on my near-endless to-do list to fix that for
Yacc++, but it isn't near the top of it.

By the way, when working with Michella Becchi on doing a hardware
regular expression engine at Intel, she studied the problem of counted
regular expressions and proposed some interesting implementation
details of how to handle them. Anyone interested in high speed regular
expression implementations would be well advised to look up her papers
on the topic.

Chris Clark email:
Compiler Resources, Inc. Web Site:
23 Bailey Rd voice: (508) 435-5016
Berlin, MA 01503 USA twitter: @intel_chris

Post a followup to this message

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