package bnfgen
Install
Dune Dependency
Authors
Maintainers
Sources
md5=f7e8200ca86a83f188d14e27644ff4a9
sha512=0606e3669cec06eb5d807b126d429346f1df7682811a3ea396b7540eaf0f8fbed4b56259089adacd848ad4d5b0a394f71aec52a34e4054d7ab2a914b065fba5d
Description
BNFGen generates random texts based on user-defined context-free grammars specified in a BNF-like syntax. There are descriptive syntax error messages and tracing options.
You can specify "weight" for rules with alternation to influence their probabilities.
For example, in <foo> ::= 10 <foo> "foo" | "foo";
the first (recursive) option will be
taken ten times more often.
You can also specify deterministic repetition ranges, like <foo>{4}
(exactly four of <foo>
)
or <foo>{1,5}
(from one to five of <foo>
).
This package includes both a library and a CLI tool based on it.
Published: 18 Feb 2021
README
BNFGen
BNFGen generates random text based on context-free grammars. You give it a file with your grammar, defined in BNF-like syntax, it gives you a string that follows that grammar.
Can be used for parser fuzzing or amusement.
An online demo is available at https://baturin.org/tools/bnfgen
Grammar syntax
The grammar syntax is an ad hoc extension of BNF (as if there's any other kind of BNF extensions).
Non-terminals are written in angle brackets (<foo>
), terminals must be quoted ('foo'
or "foo"
). It's ok to use the other kind of quotes inside quoted strings without escaping, that is, "foo 'bar' baz"
or 'foo "bar" baz'
are fine. Same kind of quote must be escaped with a backslash.
To enable writing multi-line rules easily, rules must be separated with a semicolon.
Empty alternatives are not allowed in rules with alternation, i.e. <foo> ::= <bar> | <baz>;
is fine, while <foo> ::= | <bar> | <baz>;
is not.
Here is a simple example:
$ cat examples/02_english.bnf
# A trivial subset of English.
<start> ::= <NP> <VP> ;
<NP> ::= <article> <subject> ;
<VP> ::= <verb> <article> <object> ;
<article> ::= "a" | "the" ;
<subject> ::= "customer"
| "system administrator"
| "programmer" ;
<object> ::= "system" | "compiler" | "program" ;
<verb> ::= "used" | "installed" | "developed"
$ bnfgen examples/02_english.bnf
a system administrator installed a compiler
The default start symbol is <start>
, but you can override it with --start
option. For example, bnfgen --start NP examples/02_english.bnf
will only produce noun phrases.
By default BNFGen uses space as a token separator, but you can specify any separator with --separator
option, including empty string (--separator ""
).
Repetition
There is no direct support for repetition, but it can be easily done with recursive rules.
Here are recipes for common repetition patterns:
# <foo>?
<one_or_more_foo> ::= <foo> | "" ;
# <foo>*
<zero_or_more_foo> ::= <foo> <zero_or_more_foo> | "" ;
# <foo>+
<one_or_more_foo> ::= <foo> <one_or_more_foo> | <foo> ;
BNFGen makes no attempt to check if a production will ever terminate, and even terminating rules may product a lot of data. You can limit recursion depth with --max-depth
option. Note that --max-depth 0
means "no recursion" rather than infinite. To allow infinite recursion, simply omit that option.
Weighted random
A problem with using simple BNF for producing test cases for parser fuzzing is that if each alternative is taken equally often, there's a 50% chance that production will terminate immediately with the non-recursive case. Likewise, you simply may want some symbols more frequent than others to give your text a more realistic appearance.
The "killer feature" of BNFGen is that it gives you control over recursion depth. You can specify a "weight" for every alternative to choose how often they should be taken. The default weight is 1.
For example, if you want to produce long strings of "a", you can use this rule to make BNFGen choose the recursive alternative hundred times more often.
<start> ::= 100 <start> "a" | "a" ;
You can give different weight to every alternative. For example, the following grammar will product strings with lots of a's, some b'c, and few c's.
<foo> ::= 100 "a" | 50 "b" | "c" ;
<start> ::= 100 <foo> <start> | <foo>
Installation
From the OPAM repository: opam install bnfgen
.
From a local repo clone: opam install -w .
.
opam pin add bnfgen .
You can also find some binaries in the GitHub releases.
Usage
Usage: bnfgen [OPTIONS] <BNF file>
--dump-rules Dump production rules and exit
--separator <string> Token separator for generated output, default is space
--start <string> Start symbol, default is "start"
--max-reductions <int> Maximum reductions, default is infinite
--max-nonproductive-reductions <int> Maximum number of reductions that don't produce a terminal, default is infinite
--debug Enable debug output (symbols processed, alternatives taken...)
--dump-stack Show symbol stack for every reduction (implies --debug)
--version Print version and exit
-help Display this list of options
--help Display this list of options
Library usage example
# let g = Bnfgen.grammar_from_string " <greeting> ::= \"hello\" | \"hi\" ; <start> ::= <greeting> \"world\"; " |> Result.get_ok ;;
val g : Bnfgen.Grammar.grammar =
[("greeting",
[{Bnfgen.Grammar.weight = 1; symbols = [Bnfgen.Grammar.Terminal "hi"]};
{Bnfgen.Grammar.weight = 1; symbols = [Bnfgen.Grammar.Terminal "hello"]}]);
("start",
[{Bnfgen.Grammar.weight = 1;
symbols =
[Bnfgen.Grammar.Nonterminal "greeting"; Bnfgen.Grammar.Terminal "world"]}])]
# Bnfgen.generate_string ~settings:({Bnfgen.default_settings with symbol_separator=" "}) g "start" ;;
- : (string, string) result = Ok "hello world "
# Bnfgen.generate ~settings:({Bnfgen.default_settings with symbol_separator=""}) print_endline g "start" ;;
hello world
- : (unit, string) result = Ok ()