Simple lx -l json printer
#118
Conversation
There was a problem hiding this comment.
My suggestions:
- Add json output for libfsm, not just lx. Then re(1) can use it as well as lx, and lx gets the optimisations from the codegen IR
- Escaping for special characters in json strings
- Consistency for naming (we could rename libfsm's terms if you prefer)
| #include <adt/set.h> | ||
|
|
||
| #include <fsm/fsm.h> | ||
| #include <fsm/pred.h> |
| print_zone(FILE *f, const struct ast *ast, const struct ast_zone *z) | ||
| { | ||
| struct fsm_state *s, *st; | ||
| int ret; |
|
|
||
| fprintf(f, "{\n"); | ||
| fprintf(f, " \"prefix\": { \"api\": \"%s\", \"tok\": \"%s\", \"lx\": \"%s\" },\n", | ||
| prefix.api, prefix.tok, prefix.lx); |
There was a problem hiding this comment.
I feels a bit strange to output the prefixes here - these are to affect the generated code, not to be part of it. But I know you're intending to use this json to go and generate code, so it does make sense to include them here.
| fprintf(f, " \"next_zone\": %u,\n", zindexof(ast, m->to)); | ||
| } | ||
|
|
||
| fprintf(f, " \"accepts\": true%s\n", e != NULL ? "," : ""); |
There was a problem hiding this comment.
Terminology: libfsm probably should have its .isend field renamed to "accept" at some point. I'd like the naming to be consistent.
Likewise for your "initial_zone" field, where that's called either the starting zone (when talking about the graph), or the global zone (when talking about scope).
There was a problem hiding this comment.
And for "target", which libfsm currently calls "dst" or "to".
| for (; e != NULL; e = set_next(&it)) { | ||
| struct set_iter jt; | ||
| for (st = set_first(e->sl, &jt); st != NULL; st = set_next(&jt)) { | ||
| fprintf(f, "%s\n { \"symbol\": %u, \"target\": %u }", |
There was a problem hiding this comment.
I'm not sure about these numeric symbols. Do you think a single-letter string would be more fitting?
| assert(m != NULL); | ||
|
|
||
| if (m->token != NULL) { | ||
| fprintf(f, " \"token\": \"$%s\",\n", m->token->s); |
There was a problem hiding this comment.
These strings should be escaped, per src/print/json.c's json_escputc()
| fprintf(f, " \"transitions\": ["); | ||
| for (; e != NULL; e = set_next(&it)) { | ||
| struct set_iter jt; | ||
| for (st = set_first(e->sl, &jt); st != NULL; st = set_next(&jt)) { |
There was a problem hiding this comment.
libfsm has an IR for code generation which provides a few optimisations for generated DFA. I'm not sure if you'd want to use it here. It provides different "strategies" for each state, where e.g. symbols are presented grouped as x..y ranges, or e.g. if all symbols transition to the same state. If you're going on to generate code from the json, I think it probably would make sense to make use of that.
To use that IR, I'd first add json output to libfsm (cut & pasted from the current C), which just walks through that IR and outputs what it's given. The IR was introduced here: #94
Then re -pl json xyz would give json for a single regexp, too.
|
Thanks for reviewing! Your feedback makes sense. I'm working on something else right now so I'm not sure when I'd get back to this, but I probably will eventually... 😅 For the moment, though: I didn't realize this, but I guess |
|
So they do! Replace it with a better schema if you prefer. The leaf callback is how lx renders things attached by its No worries about time! We're all busy with things. These APIs won't change, and I think this PR can wait as long as it needs. |
|
See #198 for new json for libfsm - this doesn't include the lx parts. |
I don't know what I'm doing but with enough copy-paste from other parts of the source tree I made a thing!