14

Are there any tools to convert ANTLR grammar syntax to and from other BNF syntaxes? There are several forms Backus-Naur Form (BNF, EBNF, ABNF, W3C-BNF, XBNF...) with specification, e.g. see this list. The ANTLR grammar syntax only seems to be described by examples. I know that ANTLR grammar files contain more than the specification of a context-free syntax, but you should be able to convert at least the common subset - has anyone done yet automatically?

Jakob
  • 3,570
  • 3
  • 36
  • 49

5 Answers5

12
# Grammar Syntax

|                               | BNF                           | ISO EBNF                      | ABNF                          | ANTLR                         |
|:-----------------------------:|:-----------------------------:|:-----------------------------:|:-----------------------------:|:-----------------------------:|
| rule definition               | `<name> ::= ...`              | `name = ... ;`                | `name = ...`                  | `name : ... ;`                |
| terminal items                | `...`                         | `'...'` or `"..."`            | integer or `"..."`            | `'...'`                       |
| non-terminal items            | `<...>`                       | `...`                         | `...` or `<...>`              | `...`                         |
| concatenation                 | (space)                       | `,`                           | (space)                       | (space)                       |
| choice                        | `|`                           | `|`                           | `/`                           | `|`                           |
| optional                      | requires choice syntax[^1]    | `[...]`                       | `*1...` or `[...]`            | `...?`                        |
| 0 or more repititions         | requires choice syntax[^2]    | `{...}`                       | `*...`                        | `...*`                        |
| 1 or more repititions         | requires choice syntax[^3]    | `{...}-`                      | `1*...`                       | `...+`                        |
| n repititions                 |                               | `n*...`                       | `n*n...`                      |                               |
| n to m repititions            |                               |                               | `n*m...`                      |                               |
| grouping                      |                               | `(...)`                       | `(...)`                       | `(...)`                       |
| comment                       |                               | `(*...*)`                     | `;...`                        | `// ...` or `/* ... */`       |


[^1]: `optionalb ::= a b c d | a c d`

[^2]: `list ::= | listitem list`

[^3]: `list ::= listitem | listitem list`
Travis
  • 472
  • 4
  • 10
  • Yes, you can map between different syntax at least for simple grammars. I was looking for a tool that does this conversion (BNF, ISO EBNF, ABNF, ANTLR...). Sure it is possible to create your own tool ;-) – Jakob Apr 24 '12 at 19:07
5

Jakob wrote:

The ANTLR grammar syntax only seems to be described by examples.

ANTLR (v3) is written "in its own words" (as Terence Parr himself put it) in this grammar:

http://www.antlr.org/grammar/ANTLR/ANTLRv3.g


Jakob wrote:

but you should be able to convert at least the common subset - has anyone done yet automatically?

Not that I know of. And if it does exist, I've never seen this tool being discussed on the ANTLR mailing list that I read on a regular basis.

Also note that many BNF-variants allow for left-recursive rules, something that an LL-parser generator like ANTLR cannot cope with. The left recursive rules can of course be re-factored out by the tool, but that could be rather tricky, and will probably result in a far less "readable" grammar than one would get than doing this manually.

As to converting ANTLR grammars into BNF-like form would be easier I guess, although only with the most trivial grammars. As soon as various types of predicates are put into an ANTLR grammar, the conversion might again become tricky.

Bart Kiers
  • 166,582
  • 36
  • 299
  • 288
  • Thanks, ANTLR grammar written in ANTLR was one thing I was looking for. As far as I understand ANTLR you *can* write down left-recursive grammars, but you must then refactor them by hand. A converter from other BNF-forms and ANTLR would at least reduce the need to manually adopting the differences in punctation and other minor aspects of BNF-syntax variants. – Jakob Feb 02 '11 at 10:18
  • Looks like ANTLR cannot cope with Unicode characters above Basic Multilingual Plane: ESC: '\\' ( ... | 'u' XDIGIT XDIGIT XDIGIT XDIGIT ) cannot match codepoints above \uFFFF. – Jakob Feb 02 '11 at 10:23
5

I wrote a translator for this purpose. Universal-transpiler is able to convert ANTLR grammars into PEG.js, nearley, ABNF, XBNF, and several other grammar notations. It is not yet able to translate ANTLR into W3C-BNF, but I will try to add this feature in a future version.

This translator is only compatible with a small subset of the ANTLR language, but I hope it will still be useful.

Anderson Green
  • 30,230
  • 67
  • 195
  • 328
2

There's a site that host a huge variety of grammars and tools to convert between their formats:

http://slebok.github.io/zoo/index.html

Ira Baxter
  • 93,541
  • 22
  • 172
  • 341
1

Antlr4 does allow left recursion, with remarkable flexibility. I found a modern tool to convert to and from Antlr grammars to other types of grammars. This is well supported at this time: trconvert

Frank Hileman
  • 1,159
  • 9
  • 19