1 # This file is part of NIT ( http://www.nitlanguage.org ).
3 # Licensed under the Apache License, Version 2.0 (the "License");
4 # you may not use this file except in compliance with the License.
5 # You may obtain a copy of the License at
7 # http://www.apache.org/licenses/LICENSE-2.0
9 # Unless required by applicable law or agreed to in writing, software
10 # distributed under the License is distributed on an "AS IS" BASIS,
11 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 # See the License for the specific language governing permissions and
13 # limitations under the License.
15 # Search things from the Model
17 # ModelIndex allows you to index mentities then retrieve them by their `name`
19 # It offers a set of `find_*` methods that can be used to match queries
20 # against entities name.
22 # Because each use is different, ModelIndex only provide base raw search services.
23 # All of them return IndexMatches that can be ordered and filtered by the client.
25 # ## Indexing mentities
27 # Before searching something from the ModelIndex, you have to index it.
28 # Use the `ModelIndex::index` method to do that:
31 # var index = new ModelIndex
32 # for mentity in model.collect_mentities do
33 # index.index(mentity)
39 # You can then run queries on the ModelIndex:
42 # for res in index.find("Array").limit(10) do
49 # Here some examples of how you can use the ModelIndex.
51 # ### Smart type prediction
53 # Use ModelIndex to implement a smart prediction system based on the typed prefix:
56 # var index = new ModelIndex
57 # for mentity in model.collect_mentities do
58 # # We don't really care about definitions
59 # if mentity isa MClassDef or mentity isa MPropDef then continue
60 # index.index(mentity)
63 # var typed_prefix = "Arr"
64 # for res in index.find_by_name_prefix(typed_prefix).
65 # uniq. # keep only the best ranked mentity
66 # limit(5). # limit to ten results
67 # sort(new VisibilityComparator, new NameComparator) do # order by visibility then name
72 # Will output something like:
82 # ### Method autocompletion
84 # Find methods from a class full_name:
87 # var class_full_name = "core::Array"
88 # for res in index.find_by_full_name_prefix("{class_full_name}::").
89 # uniq. # keep only the best ranked mentity
90 # sort(new VisibilityComparator). # put private in the bottom of the list
91 # limit(5). # limit to ten results
92 # sort(new FullNameComparator) do # order by lexicographic order
97 # Will output something like:
107 # ### Name typo detection and suggestion
109 # Detect unknown names and suggest corrections:
114 # if index.find_by_name_prefix(name).is_empty then
115 # printn "`{name}` not found, did you mean: "
116 # printn index.find_by_name_similarity(name).sort(new ScoreComparator).limit(2).join(" or ")
121 # Will output something like:
124 # `Zrray` not found, did you mean: Array (1) or array (1)?
128 import model
::model_collect
133 # Keep a direct link to mentities by full name to speed up `mentity_from_uri`
134 var mentities_by_full_name
: HashMap[String, MEntity] is lazy
do
135 var mentities_by_full_name
= new HashMap[String, MEntity]
136 for mentity
in collect_mentities
do
137 mentities_by_full_name
[mentity
.full_name
] = mentity
139 return mentities_by_full_name
142 # ModelIndex used to perform searches
143 var index
: ModelIndex is lazy
do
144 var index
= new ModelIndex
145 for mentity
in collect_mentities
do
146 if mentity
isa MClassDef or mentity
isa MPropDef then continue
152 redef fun mentities_by_name
(name
, filter
) do
153 var res
= new Array[MEntity]
154 if index
.name_prefixes
.has_key
(name
) then
155 for mentity
in index
.name_prefixes
[name
] do
156 if filter
== null or filter
.accept_mentity
(mentity
) then
164 redef fun mentity_by_full_name
(full_name
, filter
) do
165 if mentities_by_full_name
.has_key
(full_name
) then
166 var mentity
= mentities_by_full_name
[full_name
]
167 if filter
== null or filter
.accept_mentity
(mentity
) then return mentity
172 private var score_sorter
= new ScoreComparator
173 private var vis_sorter
= new VisibilityComparator
174 private var kind_sorter
= new MEntityComparator
175 private var name_sorter
= new NameComparator
176 private var lname_sorter
= new NameLengthComparator
177 private var fname_sorter
= new FullNameComparator
178 private var lfname_sorter
= new FullNameLengthComparator
180 # Search mentities based on a `query` string
182 # Lookup the index for anything matching `query` and return `limit` results.
184 # The algorithm used is the following:
185 # 1- lookup by name prefix
186 # 2- lookup by full_name prefix
187 # 3- loopup by levenshtein distance
189 # At each step if the `limit` is reached, the algorithm stops and returns the results.
190 fun find
(query
: String, limit
: nullable Int): Array[MEntity] do
191 # Find, lookup by name prefix
192 var matches
= index
.find_by_name_prefix
(query
).uniq
.
193 sort
(lname_sorter
, name_sorter
, kind_sorter
)
194 if limit
!= null and matches
.length
>= limit
then
195 return matches
.limit
(limit
).rerank
.sort
(vis_sorter
, score_sorter
).mentities
197 matches
= matches
.rerank
.sort
(vis_sorter
, score_sorter
)
199 # If limit not reached, lookup by full_name prefix
200 var malus
= matches
.length
201 var full_matches
= new IndexMatches
202 for match
in index
.find_by_full_name_prefix
(query
).
203 sort
(kind_sorter
, lfname_sorter
, fname_sorter
) do
205 full_matches
.add match
207 matches
= matches
.uniq
208 if limit
!= null and matches
.length
+ full_matches
.length
>= limit
then
209 matches
.add_all full_matches
210 matches
= matches
.uniq
.limit
(limit
).rerank
.sort
(vis_sorter
, score_sorter
)
211 return matches
.mentities
214 # If limit not reached, lookup by similarity
215 malus
= matches
.length
216 var sim_matches
= new IndexMatches
217 for match
in index
.find_by_similarity
(query
).sort
(score_sorter
, kind_sorter
, lname_sorter
, name_sorter
) do
219 sim_matches
.add match
221 matches
.add_all sim_matches
222 matches
= matches
.uniq
223 if limit
!= null then matches
= matches
.limit
(limit
)
224 return matches
.rerank
.sort
(vis_sorter
, score_sorter
).mentities
228 # ModelIndex indexes mentities by their name and full name
230 # It provides methods to find mentities based on a prefix or string similarity.
234 # var index = new ModelIndex
235 # for mentity in model.collect_mentities do
236 # if mentity isa MClassDef or mentity isa MPropDef then continue
237 # index.index(mentity)
240 # for e in index.find("Foo").uniq.sort(new ScoreComparator).limit(10) do
241 # print " * {e.score}: {e.mentity.name} ({e.mentity.full_name})"
246 # List of all indexed mentities.
248 # Faster than traversing the tries.
249 var mentities
= new Array[MEntity]
251 # Prefix tree for mentities `name`
253 # Because multiple mentities can share the same `name`, we use a Trie of
254 # arrays of mentities.
256 # As for now, we do not index class and property definitions.
257 # TODO add an option.
258 var name_prefixes
= new Trie[Array[MEntity]]
260 # Prefix tree for mentities `full_name`
262 # Even if two mentities cannot share the same `full_name`, we use a Trie of
263 # arrays of mentities to be consistent with `name_prefixes`.
264 var full_name_prefixes
= new Trie[Array[MEntity]]
266 # Index `mentity` by it's `MEntity::name`
268 # See `name_prefixes`.
269 private fun index_by_name
(mentity
: MEntity) do
270 var name
= mentity
.name
271 if not name_prefixes
.has_key
(name
) then
272 name_prefixes
[name
] = new Array[MEntity]
274 name_prefixes
[name
].add mentity
277 # Index `mentity` by its `MEntity::full_name`
278 private fun index_by_full_name
(mentity
: MEntity) do
279 var name
= mentity
.full_name
280 if not full_name_prefixes
.has_key
(name
) then
281 full_name_prefixes
[name
] = new Array[MEntity]
283 full_name_prefixes
[name
].add mentity
286 # Index `mentity` so it can be retrieved by a find query
288 # MEntities are indexed by both name and full_name.
289 fun index
(mentity
: MEntity) do
290 mentities
.add mentity
291 index_by_name mentity
292 index_by_full_name mentity
295 # Translate Trie results to `SearchResult`
297 # This method is used internally to translate each mentity returned by a prefix
298 # match in a Trie into a SearchResult that can be ranked by score.
300 # Results from the Trie are returned in a breadth first manner so we get the
301 # matches ordered by prefix.
302 # We preserve that order by giving an incremental score to the `array` items.
303 private fun score_results_incremental
(array
: Array[Array[MEntity]]): IndexMatches do
304 var results
= new IndexMatches
306 for mentities
in array
do
307 for mentity
in mentities
do
308 results
.add
new IndexMatch(mentity
, score
)
315 # Find all mentities where `MEntity::name` matches the `prefix`
316 fun find_by_name_prefix
(prefix
: String): IndexMatches do
317 return score_results_incremental
(name_prefixes
.find_by_prefix
(prefix
))
320 # Find all mentities where `MEntity::full_name` matches the `prefix`
321 fun find_by_full_name_prefix
(prefix
: String): IndexMatches do
322 return score_results_incremental
(full_name_prefixes
.find_by_prefix
(prefix
))
325 # Rank all mentities by the distance between `MEntity::name` and `name`
327 # Use the Levenshtein algorithm on all the indexed mentities `name`.
328 # Warning: may not scale to large indexes.
329 fun find_by_name_similarity
(name
: String): IndexMatches do
330 var results
= new IndexMatches
331 for mentity
in mentities
do
332 if mentity
isa MClassDef or mentity
isa MPropDef then continue
333 results
.add
new IndexMatch(mentity
, name
.levenshtein_distance
(mentity
.name
))
338 # Rank all mentities by the distance between `MEntity::full_name` and `full_name`
340 # Use the Levenshtein algorithm on all the indexed mentities `full_name`.
341 # Warning: may not scale to large indexes.
342 fun find_by_full_name_similarity
(name
: String): IndexMatches do
343 var results
= new IndexMatches
344 for mentity
in mentities
do
345 if mentity
isa MClassDef or mentity
isa MPropDef then continue
346 results
.add
new IndexMatch(mentity
, name
.levenshtein_distance
(mentity
.full_name
))
351 # Rank all mentities by the distance between `name` and both the mentity name and full name
352 fun find_by_similarity
(name
: String): IndexMatches do
353 var results
= new IndexMatches
354 for mentity
in mentities
do
355 if mentity
isa MClassDef or mentity
isa MPropDef then continue
356 results
.add
new IndexMatch(mentity
, name
.levenshtein_distance
(mentity
.name
))
357 results
.add
new IndexMatch(mentity
, name
.levenshtein_distance
(mentity
.full_name
))
362 # Find mentities by name trying first by prefix then by similarity
363 fun find_by_name
(name
: String): IndexMatches do
364 var results
= find_by_name_prefix
(name
)
365 for mentity
in mentities
do
366 if mentity
isa MClassDef or mentity
isa MPropDef then continue
367 results
.add
new IndexMatch(mentity
, name
.levenshtein_distance
(mentity
.name
))
372 # Find mentities by full name trying firt by prefix then by similarity
373 fun find_by_full_name
(name
: String): IndexMatches do
374 var results
= find_by_full_name_prefix
(name
)
375 for mentity
in mentities
do
376 if mentity
isa MClassDef or mentity
isa MPropDef then continue
377 results
.add
new IndexMatch(mentity
, name
.levenshtein_distance
(mentity
.full_name
))
382 # Find all mentities that matches `name`
384 # 1. try by name prefix
385 # 2. add full name prefix matches
386 # 3. try similarity by name
387 # 4. try similarity by full_name
388 fun find
(name
: String): IndexMatches do
389 var results
= find_by_name_prefix
(name
)
391 for result
in find_by_full_name_prefix
(name
) do
395 for mentity
in mentities
do
396 if mentity
isa MClassDef or mentity
isa MPropDef then continue
397 results
.add
new IndexMatch(mentity
, name
.levenshtein_distance
(mentity
.name
))
398 results
.add
new IndexMatch(mentity
, name
.levenshtein_distance
(mentity
.full_name
))
404 # An array of IndexMatch instances returned by the ModelIndex
406 # Index matches can be sorted, filtered and truncated.
408 # Thanks to the fluent interface, the index matches can be manipulated in chain
409 # from a model index query:
412 # var res = index.find("Foo").
414 # sort(new ScoreComparator, new MEntityComparator).
416 # sort(new VisibilityComparator)
419 super Array[IndexMatch]
421 # Create a new ModelMatches from an array of matches
423 # Elements are copied.
424 init from_matches
(matches
: Array[IndexMatch]) do self.add_all matches
426 # Sort the matches with `comparator` (or a list of comparators)
428 # Return a new IndexMatches instance with the sorted results.
430 # When more than one comparator is given, the comparators are applied in a
431 # pipeline where the `n`th comparator is applied only if the `n-1`th comparator
433 fun sort
(comparator
: ScoreComparator...): IndexMatches do
435 if comparator
.length
== 1 then
436 comparator
.first
.sort res
438 var comparators
= new MatchComparators(comparator
)
441 return new IndexMatches.from_matches
(res
)
444 # Limit the matches with `limit`
446 # Return a new IndexMatches instance with only the `limit` first matches.
447 fun limit
(limit
: Int): IndexMatches do
448 var res
= new Array[IndexMatch]
450 if res
.length
>= limit
then break
453 return new IndexMatches.from_matches
(res
)
456 # Remove doublons from the matches
458 # Preverse the lowest score of all the matches for a MEntity.
459 fun uniq
: IndexMatches do
460 var scores
= new HashMap[MEntity, IndexMatch]
461 var res
= new Array[IndexMatch]
463 var mentity
= match
.mentity
464 if scores
.has_key
(mentity
) then
465 var older
= scores
[mentity
]
466 if match
.score
< older
.score
then older
.score
= match
.score
468 scores
[mentity
] = match
472 return new IndexMatches.from_matches
(res
)
475 # Reset score of each matches to follow `self` order
477 # Usefull when you need to apply one sorter over another.
478 fun rerank
: IndexMatches do
479 var res
= new IndexMatches
482 match
.score
= res
.length
487 # Aggregate the mentities for all the matches
489 # Preserve the match order.
490 fun mentities
: Array[MEntity] do
491 var res
= new Array[MEntity]
492 for match
in self do res
.add match
.mentity
497 # An MEntity matched from a ModelIndex
499 # Each match has a `score`. The score should be seen as the distance of
500 # the match from the query. In other words, the lowest is the score, the more
501 # relevant is the match.
505 redef type OTHER: IndexMatch
510 # Score allocated by the search method
512 # A lowest score means a more relevant match.
514 # Scores values are arbitrary, the meaning of `10` vs `2000` really depends
515 # on the search method producing the match and the comparators used to sort
517 # The only universal rule is: low score = relevance.
518 var score
: Int is writable
520 # By default matches are compared only on their score
521 redef fun <=>(o
) do return score
<=> o
.score
523 redef fun to_s
do return "{mentity} ({score})"
526 # Compare two matches by their score
528 # Since the score can be seen as a distance, we want the lowest score first.
529 class ScoreComparator
532 redef type COMPARED: IndexMatch
534 redef fun compare
(o1
, o2
) do return o1
.score
<=> o2
.score
537 # A pipeline of comparators executed in inclusion order
539 # This class is used internally to agregate the behaviors of multiple comparators.
540 # Use `IndexMatches::sort(comparator...)` instead.
541 private class MatchComparators
542 super ScoreComparator
544 # Comparator to use in the array order
545 var comparators
: Array[ScoreComparator]
547 # Compare with each comparator
549 # Return the compare value of the first one to return anything else than 0.
550 redef fun compare
(o1
, o2
) do
551 for comparator
in comparators
do
552 var c
= comparator
.compare
(o1
, o2
)
553 if c
!= 0 then return c
559 # Compare two matches by their MEntity kind
561 # Usefull to order the mentities by kind in this order:
562 # packages, groups, modules and classes, properties.
563 class MEntityComparator
564 super ScoreComparator
566 # See `MEntity::compare_mentity`
567 redef fun compare
(o1
, o2
) do
568 return o1
.mentity
.mentity_kind_rank
<=> o2
.mentity
.mentity_kind_rank
572 # Compare two matches by their MEntity visibility
574 # We reverse the original order so private is at the end of the list.
575 class VisibilityComparator
576 super ScoreComparator
578 redef fun compare
(o1
, o2
) do return o2
.mentity
.visibility
<=> o1
.mentity
.visibility
581 # Compare two matches by their name in lexicographic order
583 # Generally, for a same score, we want to put A before Z.
585 super ScoreComparator
587 redef fun compare
(o1
, o2
) do return o1
.mentity
.name
<=> o2
.mentity
.name
590 # Compare two matches by their name length
591 class NameLengthComparator
592 super ScoreComparator
594 redef fun compare
(o1
, o2
) do return o1
.mentity
.name
.length
<=> o2
.mentity
.name
.length
597 # Compare two matches by their full_name in lexicographic order
599 # Generally, for a same score, we want to put A before Z.
600 class FullNameComparator
601 super ScoreComparator
603 redef fun compare
(o1
, o2
) do return o1
.mentity
.full_name
<=> o2
.mentity
.full_name
606 # Compare two matches by their full name length
607 class FullNameLengthComparator
608 super ScoreComparator
610 redef fun compare
(o1
, o2
) do return o1
.mentity
.full_name
.length
<=> o2
.mentity
.full_name
.length
615 # Compare MEntity class kind
617 # Unknown kind have a virtually high score.
618 private fun mentity_kind_rank
: Int do return 10
622 redef fun mentity_kind_rank
do return 1
626 redef fun mentity_kind_rank
do return 2
630 redef fun mentity_kind_rank
do return 3
634 redef fun mentity_kind_rank
do return 4
637 redef class MClassDef
638 redef fun mentity_kind_rank
do return 5
641 redef class MProperty
642 redef fun mentity_kind_rank
do return 6
646 redef fun mentity_kind_rank
do return 7