Class: Mongrel::URIClassifier

Inherits:
Object
  • Object
show all
Defined in:
lib/mongrel.rb,
ext/http11/http11.c

Instance Method Summary collapse

Constructor Details

#new ⇒ URIClassifier

Initializes a new URIClassifier object that you can use to associate URI sequences with objects. You can actually use it with any string sequence and any objects, but it’s mostly used with URIs.

It uses TST from www.octavian.org/cs/software.html to build an ternary search trie to hold all of the URIs. It uses this to do an initial search for the a URI prefix, and then to break the URI into SCRIPT_NAME and PATH_INFO portions. It actually will do two searches most of the time in order to find the right handler for the registered prefix portion.



355
356
357
358
359
360
361
362
363
364
 
# File 'ext/http11/http11.c', line 355

VALUE URIClassifier_init(VALUE self)
{
  VALUE hash;

  // we create an internal hash to protect stuff from the GC
  hash = rb_hash_new();
  rb_ivar_set(self, id_handler_map, hash);

  return self;
}

Instance Method Details

#inspect ⇒ Object

Simply does an inspect that looks like a Hash inspect.



26
27
28
 
# File 'lib/mongrel.rb', line 26

def inspect
  @handler_map.inspect
end

#register("/someuri", SampleHandler.new) ⇒ nil

Registers the SampleHandler (one for all requests) with the “/someuri”. When URIClassifier::resolve is called with “/someuri” it’ll return SampleHandler immediately. When called with “/someuri/iwant” it’ll also return SomeHandler immediatly, with no additional searches, but it will return path info with “/iwant”.

You actually can reuse this class to register nearly anything and quickly resolve it. This could be used for caching, fast mapping, etc. The downside is it uses much more memory than a Hash, but it can be a lot faster. It’s main advantage is that it works on prefixes, which is damn hard to get right with a Hash.

Returns:

  • (nil) 


383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
 
# File 'ext/http11/http11.c', line 383

VALUE URIClassifier_register(VALUE self, VALUE uri, VALUE handler)
{
  int rc = 0;
  void *ptr = NULL;
  struct tst *tst = NULL;
  DATA_GET(self, struct tst, tst);

  rc = tst_insert((unsigned char *)StringValueCStr(uri), (void *)handler , tst, 0, &ptr);

  if(rc == TST_DUPLICATE_KEY) {
    rb_raise(rb_eStandardError, "Handler already registered with that name");
  } else if(rc == TST_ERROR) {
    rb_raise(rb_eStandardError, "Memory error registering handler");
  } else if(rc == TST_NULL_KEY) {
    rb_raise(rb_eStandardError, "URI was empty");
  }
  
  rb_hash_aset(rb_ivar_get(self, id_handler_map), uri, handler);

  return Qnil;
}

#resolve("/someuri") ⇒ Object #resolve("/someuri/pathinfo") ⇒ Object #resolve("/notfound/orhere") ⇒ nil #resolve("/") ⇒ Object #resolve("/path/from/root") ⇒ Object

Attempts to resolve either the whole URI or at the longest prefix, returning the prefix (as script_info), path (as path_info), and registered handler (usually an HttpHandler). If it doesn’t find a handler registered at the longest match then it returns nil,nil,nil.

Because the resolver uses a trie you are able to register a handler at any character in the URI and it will be handled as long as it’s the longest prefix. So, if you registered handler #1 at “/something/lik”, and #2 at “/something/like/that”, then a a search for “/something/like” would give you #1. A search for “/something/like/that/too” would give you #2.

This is very powerful since it means you can also attach handlers to parts of the ; (semi-colon) separated path params, any part of the path, use off chars, anything really. It also means that it’s very efficient to do this only taking as long as the URI has characters.

A slight modification to the CGI 1.2 standard is given for handlers registered to “/”. CGI expects all CGI scripts to be at some script path, so it doesn’t really say anything about a script that handles the root. To make this work, the resolver will detect that the requested handler is at “/”, and return that for script_name, and then simply return the full URI back as path_info.

It expects strings with no embedded ‘0’ characters. Don’t try other string-like stuff yet.

Overloads:

  • #resolve("/notfound/orhere") ⇒ nil

    Returns:

    • (nil, nil, nil) 


462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
 
# File 'ext/http11/http11.c', line 462

VALUE URIClassifier_resolve(VALUE self, VALUE uri)
{
  void *handler = NULL;
  int pref_len = 0;
  struct tst *tst = NULL;
  VALUE result;
  unsigned char *uri_str = NULL;

  DATA_GET(self, struct tst, tst);
  uri_str = (unsigned char *)StringValueCStr(uri);

  handler = tst_search(uri_str, tst, &pref_len);

  // setup for multiple return values
  result = rb_ary_new();

  if(handler) {
    rb_ary_push(result, rb_str_substr (uri, 0, pref_len));
    // compensate for a script_name="/" where we need to add the "/" to path_info to keep it consistent
    if(pref_len == 1 && uri_str[0] == '/') {
      // matches the root URI so we have to use the whole URI as the path_info
      rb_ary_push(result, uri);
    } else {
      // matches a script so process like normal
      rb_ary_push(result, rb_str_substr(uri, pref_len, RSTRING(uri)->len));
    }
      
    rb_ary_push(result, (VALUE)handler);
  } else {
    // not found so push back nothing
    rb_ary_push(result, Qnil);
    rb_ary_push(result, Qnil);
    rb_ary_push(result, Qnil);
  }

  return result;
}

#unregister("/someuri") ⇒ Object

Yep, just removes this uri and it’s handler from the trie.



412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
 
# File 'ext/http11/http11.c', line 412

VALUE URIClassifier_unregister(VALUE self, VALUE uri)
{
  void *handler = NULL;
  struct tst *tst = NULL;
  DATA_GET(self, struct tst, tst);

  handler = tst_delete((unsigned char *)StringValueCStr(uri), tst);

  if(handler) {
    rb_hash_delete(rb_ivar_get(self, id_handler_map), uri);

    return (VALUE)handler;
  } else {
    return Qnil;
  }
}

#uris ⇒ Object

Returns the URIs that have been registered with this classifier so far. The URIs returned should not be modified as this will cause a memory leak. You can use this to inspect the contents of the URIClassifier.



21
22
23
 
# File 'lib/mongrel.rb', line 21

def uris
  @handler_map.keys
end