mongodb_meilisearch 1.3.0 → 2.1.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/Gemfile.lock +5 -4
- data/README.org +648 -0
- data/lib/mongodb_meilisearch/version.rb +1 -1
- data/lib/search/class_methods.rb +25 -19
- data/lib/search/client.rb +178 -21
- data/lib/search/errors.rb +7 -0
- data/lib/search/instance_methods.rb +11 -10
- metadata +5 -4
- data/README.md +0 -531
data/README.org
ADDED
@@ -0,0 +1,648 @@
|
|
1
|
+
#+OPTIONS: toc:2
|
2
|
+
* Table of Contents :TOC:
|
3
|
+
- [[#mongodbmeilisearch][MongodbMeilisearch]]
|
4
|
+
- [[#installation][Installation]]
|
5
|
+
- [[#usage][Usage]]
|
6
|
+
- [[#model-integration][Model Integration]]
|
7
|
+
- [[#indexes][Indexes]]
|
8
|
+
- [[#searching][Searching]]
|
9
|
+
- [[#upgrading][Upgrading]]
|
10
|
+
- [[#development][Development]]
|
11
|
+
- [[#license][License]]
|
12
|
+
- [[#code-of-conduct][Code of Conduct]]
|
13
|
+
- [[#footnotes][Footnotes]]
|
14
|
+
|
15
|
+
* MongodbMeilisearch
|
16
|
+
A simple gem for integrating
|
17
|
+
[[https://www.meilisearch.com][Meilisearch]] into Ruby applications
|
18
|
+
that are backed by [[https://www.mongodb.com/][MongoDB]].
|
19
|
+
|
20
|
+
** Installation
|
21
|
+
Install the gem and add to the application's Gemfile by executing:
|
22
|
+
|
23
|
+
#+begin_example
|
24
|
+
$ bundle add mongodb_meilisearch
|
25
|
+
#+end_example
|
26
|
+
|
27
|
+
If bundler is not being used to manage dependencies, install the gem by
|
28
|
+
executing:
|
29
|
+
|
30
|
+
#+begin_example
|
31
|
+
$ gem install mongodb_meilisearch
|
32
|
+
#+end_example
|
33
|
+
|
34
|
+
** Usage
|
35
|
+
A high level overview
|
36
|
+
|
37
|
+
*** Pre-Requisites
|
38
|
+
- [[https://www.meilisearch.com][Meilisearch]]
|
39
|
+
- [[https://www.mongodb.com/][MongoDB]]
|
40
|
+
- Some models that =include Mongoid::Document=
|
41
|
+
|
42
|
+
*** Configuration
|
43
|
+
**** Background
|
44
|
+
Meilisearch uses 3 different API keys.
|
45
|
+
- A Master API Key
|
46
|
+
- An Administrative API key
|
47
|
+
- A Search API key
|
48
|
+
|
49
|
+
The Master API key should be used when starting Meilisearch,[fn:macos] but should /never/ be used when searching or administering [[https://www.meilisearch.com/docs/learn/security/basic_security][for security reasons]].
|
50
|
+
|
51
|
+
When you start it with a master key, it will create default Search & Administrative API Keys.[fn:default_keys] These ~mongodb_meilisearch~ uses these for searching & administering respectively.
|
52
|
+
|
53
|
+
**** Details
|
54
|
+
=MEILI_MASTER_KEY= is the environment variable Meilisearch uses for setting its master key at startup if it isn't provided via command line arguments.
|
55
|
+
|
56
|
+
=MEILISEARCH_SEARCH_KEY= & =MEILISEARCH_ADMIN_KEY= are for Search & administration respectively. Administration is /everything/ that isn't search, including things like adding new records to the index.
|
57
|
+
|
58
|
+
I recommend you define the Search & Administration keys. If you provide both of those the =MEILI_MASTER_KEY= will /not/ be required, /or used/ if present. If you only know the master key, there are details below for how to obtain the others.
|
59
|
+
|
60
|
+
Define the following variables in your environment (or =.env= file if
|
61
|
+
you're using [[https://github.com/bkeepers/dotenv][dotenv]]). The url below is the default one Meilisearch
|
62
|
+
uses when run locally. Change it if you're running it on a different box or have otherwise changed the default location setup.
|
63
|
+
|
64
|
+
#+begin_src sh
|
65
|
+
SEARCH_ENABLED=true
|
66
|
+
MEILI_MASTER_KEY=<your search master key>
|
67
|
+
MEILISEARCH_SEARCH_KEY=<your search api key here>
|
68
|
+
MEILISEARCH_ADMIN_KEY=<your admin api key here>
|
69
|
+
MEILISEARCH_URL=http://127.0.0.1:7700
|
70
|
+
#+end_src
|
71
|
+
|
72
|
+
Optional configuration
|
73
|
+
|
74
|
+
#+begin_src sh
|
75
|
+
MEILISEARCH_TIMEOUT=10
|
76
|
+
MEILISEARCH_MAX_RETRIES=2
|
77
|
+
#+end_src
|
78
|
+
|
79
|
+
|
80
|
+
👋 Environment variables starting with =MEILI_= are [[https://www.meilisearch.com/docs/learn/configuration/instance_options][ones defined by Meilisearch]]. The ones starting with =MEILISEARCH_= are specific to this library.
|
81
|
+
|
82
|
+
***** Finding your Search & Admin keys
|
83
|
+
Meilisearch doesn't make it easy to find out what your default search & administration API keys are, and I haven't found a way to define them yourself.
|
84
|
+
|
85
|
+
Fortunately for you, this library has an easy way, and you've got this library. 😉
|
86
|
+
|
87
|
+
#+begin_src ruby
|
88
|
+
Search::Client.instance.get_default_keys("your master key here")
|
89
|
+
#+end_src
|
90
|
+
|
91
|
+
This will return a hash with the keys you'll need.
|
92
|
+
|
93
|
+
#+begin_src ruby
|
94
|
+
{:search=>"your search api key",
|
95
|
+
:admin=>"your admin api key"}
|
96
|
+
#+end_src
|
97
|
+
|
98
|
+
Use those keys in your environment. The next time the =Search::Client= is initialized it'll use them.
|
99
|
+
|
100
|
+
***** Validating Search & Admin keys
|
101
|
+
When switching between Meilisearch instances the same master key doesn't result in the same default admin and search keys. Being able to easily validate this is useful in situations like testing code against Meilisearch instances that are inside a Docker container /and/ in the Host OS.
|
102
|
+
|
103
|
+
Practically speaking you may want to simply /not/ provide these keys during development, and only use the =MEILI_MASTER_KEY=. Just be sure to include them when talking to a production instance.
|
104
|
+
|
105
|
+
#+begin_src ruby
|
106
|
+
Search::Client.instance.validate_default_keys(ENV['MEILI_MASTER_KEY'])
|
107
|
+
#+end_src
|
108
|
+
|
109
|
+
This returns a hash like the following.
|
110
|
+
|
111
|
+
- =status= is always either ="provided"= or ="missing"= and indicates if the =MEILISEARCH_SEARCH_KEY= or =MEILISEARCH_ADMIN_KEY= were found in the environment.
|
112
|
+
- =matches= is always either =true= or =false=, and indicates if the provided Search & Admin keys (if any) match the defaults the current instance is returning.
|
113
|
+
|
114
|
+
#+begin_src ruby
|
115
|
+
{
|
116
|
+
:search_key=>{
|
117
|
+
:status=>"provided",
|
118
|
+
:matches=>true
|
119
|
+
},
|
120
|
+
:admin_key=>{
|
121
|
+
:status=>"provided",
|
122
|
+
:matches=>true
|
123
|
+
}
|
124
|
+
}
|
125
|
+
#+end_src
|
126
|
+
|
127
|
+
** Model Integration
|
128
|
+
Add the following near the top of your model. Only the =extend= and
|
129
|
+
=include= lines are required. This assumes your model also includes
|
130
|
+
=Mongoid::Document=
|
131
|
+
|
132
|
+
#+begin_src ruby
|
133
|
+
include Search::InstanceMethods
|
134
|
+
extend Search::ClassMethods
|
135
|
+
#+end_src
|
136
|
+
|
137
|
+
If you want Rails to automatically add, update, and delete records from
|
138
|
+
the index, add the following to your model.
|
139
|
+
|
140
|
+
You can override these methods if needed, but you're unlikely to want
|
141
|
+
to.
|
142
|
+
|
143
|
+
#+begin_src ruby
|
144
|
+
# enabled?() is controlled by the SEARCH_ENABLED environment variable
|
145
|
+
if Search::Client.instance.enabled?
|
146
|
+
after_create :add_to_search
|
147
|
+
after_update :update_in_search
|
148
|
+
after_destroy :remove_from_search
|
149
|
+
end
|
150
|
+
#+end_src
|
151
|
+
|
152
|
+
Assuming you've done the above a new index will be created with a name
|
153
|
+
that corresponds to your model's name, only in snake case. All of your
|
154
|
+
models fields will be indexed and
|
155
|
+
[[https://www.meilisearch.com/docs/learn/fine_tuning_results/filtering][filterable]].
|
156
|
+
|
157
|
+
*** Example Rails Model
|
158
|
+
Here's what it looks like when you put it all together in a Rails model
|
159
|
+
with the default behavior.
|
160
|
+
|
161
|
+
#+begin_src ruby
|
162
|
+
class Person
|
163
|
+
include Mongoid::Document
|
164
|
+
extend Search::ClassMethods
|
165
|
+
|
166
|
+
if Search::Client.instance.enabled?
|
167
|
+
after_create :add_to_search
|
168
|
+
after_update :update_in_search
|
169
|
+
after_destroy :remove_from_search
|
170
|
+
end
|
171
|
+
|
172
|
+
# normal Mongoid attributes
|
173
|
+
field :name, type: String
|
174
|
+
field :description, type: String
|
175
|
+
field :age, type: Integer
|
176
|
+
end
|
177
|
+
#+end_src
|
178
|
+
|
179
|
+
Note that that /unless you configure it otherwise/ the ids of
|
180
|
+
=belongs_to= objects will not be searchable. This is because they're
|
181
|
+
random strings that no human's ever going to be searching for, and we
|
182
|
+
don't want to waste RAM or storage.
|
183
|
+
|
184
|
+
*** Going Beyond The Defaults
|
185
|
+
This module strives for sensible defaults, but you can override them
|
186
|
+
with the following optional constants:
|
187
|
+
|
188
|
+
- =PRIMARY_SEARCH_KEY= - a Symbol matching one of your model's
|
189
|
+
attributes that is guaranteed unique. This defaults to =_id=
|
190
|
+
- =SEARCH_INDEX_NAME= - a String - useful if you want to have records
|
191
|
+
from multiple classes come back in the same search results. This
|
192
|
+
defaults to the underscored form of the current class name.
|
193
|
+
- =SEARCH_OPTIONS= - a hash of key value pairs in JS style
|
194
|
+
- See the
|
195
|
+
[[https://www.meilisearch.com/docs/reference/api/search#search-parameters][meilisearch search parameter docs]] for details.
|
196
|
+
- example from
|
197
|
+
[[https://github.com/meilisearch/meilisearch-ruby/blob/main/spec/meilisearch/index/search/multi_params_spec.rb][meliesearch's multi_param_spec]]
|
198
|
+
|
199
|
+
#+begin_src ruby
|
200
|
+
{
|
201
|
+
attributesToCrop: ['title'],
|
202
|
+
cropLength: 2,
|
203
|
+
filter: 'genre = adventure',
|
204
|
+
attributesToHighlight: ['title'],
|
205
|
+
limit: 2
|
206
|
+
}
|
207
|
+
#+end_src
|
208
|
+
|
209
|
+
- =SEARCH_RANKING_RULES= - an array of strings that correspond to
|
210
|
+
meilisearch rules see
|
211
|
+
[[https://www.meilisearch.com/docs/learn/core_concepts/relevancy#ranking-rules][meilisearch ranking rules docs]] You probably don't want to change this.
|
212
|
+
|
213
|
+
** Indexes
|
214
|
+
Searching is limited to records that have been added to a given index.
|
215
|
+
This means, if you want to perform one search and get back records from
|
216
|
+
multiple models you'll need to add them to the same index.
|
217
|
+
|
218
|
+
In order to do that add the =SEARCH_INDEX_NAME= constant to the model
|
219
|
+
whose search stuff you want to end up in the same index. You can name
|
220
|
+
this just about anything. The important thing is that all the models
|
221
|
+
that share this index have the same =SEARCH_INDEX_NAME= constant
|
222
|
+
defined. You may want to just add it to a module they all import.
|
223
|
+
|
224
|
+
#+begin_src ruby
|
225
|
+
SEARCH_INDEX_NAME='general_search'
|
226
|
+
#+end_src
|
227
|
+
|
228
|
+
If multiple models are using the same index, you should also add
|
229
|
+
=CLASS_PREFIXED_SEARCH_IDS=true=. This causes the =id= field to be
|
230
|
+
=<ClassName>_<_id>= For example, a =Note= record might have an index of
|
231
|
+
="Note_64274543906b1d7d02c1fcc6"=. If undefined this will default to
|
232
|
+
=false=. This is not needed if you can absolutely guarantee that there
|
233
|
+
will be no overlap in ids amongst all the models using a shared index.
|
234
|
+
|
235
|
+
#+begin_src ruby
|
236
|
+
CLASS_PREFIXED_SEARCH_IDS=true
|
237
|
+
#+end_src
|
238
|
+
|
239
|
+
Setting =CLASS_PREFIXED_SEARCH_IDS= to =true= will also cause the
|
240
|
+
original Mongoid =_id= field to be indexed as =original_document_id=.
|
241
|
+
This is useful if you want to be able to retrieve the original record
|
242
|
+
from the database.
|
243
|
+
|
244
|
+
*** Searchable Data
|
245
|
+
You probably don't want to index /all/ the fields. For example, unless
|
246
|
+
you intend to allow users to sort by when a record was created, there's
|
247
|
+
no point in recording it's =created_at= in the search index. It'll just
|
248
|
+
waste bandwidth, memory, and disk space.
|
249
|
+
|
250
|
+
Define a =SEARCHABLE_ATTRIBUTES= constant with an array of strings to
|
251
|
+
limit things. These are the field names, and/or names of methods you
|
252
|
+
wish to have indexed.
|
253
|
+
|
254
|
+
By default these will /also/ be the fields you can filter on.
|
255
|
+
|
256
|
+
Note that Meilisearch requires there to be an =id= field and it must be
|
257
|
+
a string. If you don't define one it will use string version of the
|
258
|
+
=_id= your document's =BSON::ObjectId=.
|
259
|
+
|
260
|
+
#+begin_src ruby
|
261
|
+
# explicitly define the fields you want to be searchable
|
262
|
+
# this should be an array of symbols
|
263
|
+
SEARCHABLE_ATTRIBUTES = %w[title body]
|
264
|
+
# OR explicitly define the fields you DON'T want searchable
|
265
|
+
SEARCHABLE_ATTRIBUTES = searchable_attributes - [:created_at]
|
266
|
+
#+end_src
|
267
|
+
|
268
|
+
**** Including Foreign Key data
|
269
|
+
If, for example, your =Person= =belongs_to: group= and you wanted that
|
270
|
+
group's id to be searchable you would include =group_id= in the list.
|
271
|
+
|
272
|
+
If you don't specify any =SEARCHABLE_ATTRIBUTES=, the default list will
|
273
|
+
exclude any fields that are =Mongoid::Fields::ForeignKey= objects.
|
274
|
+
|
275
|
+
**** Getting Extra Specific
|
276
|
+
If your searchable data needs to by dynamically generated instead of
|
277
|
+
just taken directly from the =Mongoid::Document='s attributes or
|
278
|
+
existing methods you can define a =search_indexable_hash= method on your
|
279
|
+
class.
|
280
|
+
|
281
|
+
Before you do, please note that as of v1.1 your =SEARCHABLE_ATTRIBUTES=
|
282
|
+
constant can contain fields and method names in its array of values.
|
283
|
+
Making a method for each thing dynamically generated thing you want in
|
284
|
+
the search and then including it in SEARCHABLE_ATTRIBUTES is going to be
|
285
|
+
the easiest way of accomplishing this.
|
286
|
+
|
287
|
+
Your =search_indexable_hash= must return a hash, and that hash must
|
288
|
+
include the following keys: - ="id"= - a string that uniquely identifies
|
289
|
+
the record - ="object_class"= the name of the class that this record
|
290
|
+
corresponds to.
|
291
|
+
|
292
|
+
The value of ="object_class"= is usually just =self.class.name=. This is
|
293
|
+
something specific to this gem, and not Meilisearch itself.
|
294
|
+
|
295
|
+
See =InstanceMethods#search_indexable_hash= for an example.
|
296
|
+
|
297
|
+
**** Filterable Fields
|
298
|
+
If you'd like to only be able to filter on a subset of those then you
|
299
|
+
can define =FILTERABLE_ATTRIBUTE_NAMES= but it /must/ be a subset of
|
300
|
+
=SEARCHABLE_ATTRIBUTES=. This is enforced by the gem to guarantee no
|
301
|
+
complaints from Meilisearch. These must be symbols.
|
302
|
+
|
303
|
+
If you have no direct need for filterable results, set
|
304
|
+
=UNFILTERABLE_IN_SEARCH=true= in your model. This will save on index
|
305
|
+
size and speed up indexing, but you won't be able to filter search
|
306
|
+
results, and that's half of what makes Meilisearch so great. It should
|
307
|
+
be noted, that even if this /is/ set to =true= this gem will still add
|
308
|
+
="object_class"= as a filterable attribute.
|
309
|
+
|
310
|
+
This is the magic that allows you to have an index shared by multiple
|
311
|
+
models and still be able to retrieve results specifically for one.
|
312
|
+
|
313
|
+
If you decide to re-enable filtering you can remove that constant, or
|
314
|
+
set it to false. Then call the following. If
|
315
|
+
=FILTERABLE_ATTRIBUTE_NAMES= is defined it will use that, otherwise it
|
316
|
+
will use whatever =.searchable_attributes= returns.
|
317
|
+
|
318
|
+
#+begin_src ruby
|
319
|
+
MyModel.set_filterable_attributes! # synchronous
|
320
|
+
MyModel.set_filterable_attributes # asynchronous
|
321
|
+
#+end_src
|
322
|
+
|
323
|
+
This will cause Meilisearch to reindex all the records for that index.
|
324
|
+
If you have a large number of records this could take a while. Consider
|
325
|
+
running it on a background thread. Note that filtering is managed at the
|
326
|
+
index level, not the individual record level. By setting filterable
|
327
|
+
attributes you're giving Meilisearch guidance on what to do when
|
328
|
+
indexing your data.
|
329
|
+
|
330
|
+
Note that you will encounter problems in a shared index if you try and
|
331
|
+
filter on a field that one of the contributing models doesn't have set
|
332
|
+
as a filterable field, or doesn't have at all.
|
333
|
+
|
334
|
+
*** Sortable Fields
|
335
|
+
Sortable fields work in essentially the same way as filterable fields.
|
336
|
+
By default it's the same as your =FILTERABLE_ATTRIBUTE_NAMES= which, in
|
337
|
+
turn, defaults to your =SEARCHABLE_ATTRIBUTES= You can override it by
|
338
|
+
setting =SORTABLE_ATTRIBUTE_NAMES=.
|
339
|
+
|
340
|
+
Note that you will encounter problems in a shared index if you try and
|
341
|
+
sort on a field that one of the contributing models doesn't have set as
|
342
|
+
a sortable field, or doesn't have at all.
|
343
|
+
|
344
|
+
#+begin_src ruby
|
345
|
+
MyModel.set_sortable_attributes! # synchronous
|
346
|
+
MyModel.set_sortable_attributes # asynchronous
|
347
|
+
#+end_src
|
348
|
+
|
349
|
+
*** Indexing things
|
350
|
+
*Important note*: By default anything you do that updates the search
|
351
|
+
index (adding, removing, or changing) happens asynchronously.
|
352
|
+
|
353
|
+
Sometimes, especially when debugging something on the console, you want
|
354
|
+
to update the index /synchronously/. The convention used in this
|
355
|
+
codebase - and in the meilisearch-ruby library we build on - is that the
|
356
|
+
synchronous methods are the ones with the bang. Similar to how mutating
|
357
|
+
state is potentially dangerous and noted with a bang, using synchronous
|
358
|
+
methods is potentially problematic for your users, and thus noted with a
|
359
|
+
bang.
|
360
|
+
|
361
|
+
For example:
|
362
|
+
|
363
|
+
#+begin_src ruby
|
364
|
+
MyModel.reindex # runs asyncronously
|
365
|
+
#+end_src
|
366
|
+
|
367
|
+
vs
|
368
|
+
|
369
|
+
#+begin_src ruby
|
370
|
+
MyModel.reindex! # runs synchronously
|
371
|
+
#+end_src
|
372
|
+
|
373
|
+
**** Reindexing, Adding, Updating, and Deleting
|
374
|
+
*Reindexing*
|
375
|
+
Calling =MyModel.reindex!= deletes all the existing records from the
|
376
|
+
current index, and then reindexes all the records for the current model.
|
377
|
+
It's safe to run this even if there aren't any records. In addition to
|
378
|
+
re-indexing your models, it will update/set the "sortable" and
|
379
|
+
"filterable" fields on the relevant indexes.
|
380
|
+
|
381
|
+
Note: reindexing behaves slightly differently than all the other
|
382
|
+
methods. It runs semi-asynchronously by default. The Asynchronous form
|
383
|
+
will first, attempt to /synchronously/ delete all the records from the
|
384
|
+
index. If that fails an exception will be raised. Otherwise you'd think
|
385
|
+
everything was fine when actually it had failed miserably. If you call
|
386
|
+
=.reindex!= it will be entirely synchronous.
|
387
|
+
|
388
|
+
Note: adding, updating, and deleting should happen automatically if
|
389
|
+
you've defined =after_create=, =after_update=, and =after_destroy= as
|
390
|
+
instructed above. You'll mostly only want to use these when manually
|
391
|
+
mucking with things in the console.
|
392
|
+
|
393
|
+
*Adding*
|
394
|
+
Be careful to not add documents that are already in the index.
|
395
|
+
|
396
|
+
- Add everything: =MyClass.add_all_to_search=
|
397
|
+
- Add a specific instance: =my_instance.add_to_search=
|
398
|
+
- Add a specific subset of documents:
|
399
|
+
=MyClass.add_documents(documents_hashes)= IMPORTANT:
|
400
|
+
=documents_hashes= must be an array of hashes that were each generated
|
401
|
+
via =search_indexable_hash=
|
402
|
+
|
403
|
+
*Updating*
|
404
|
+
- Update everything: call =reindex=
|
405
|
+
- Update a specific instance: =my_instance.update_in_search=
|
406
|
+
- Update a specific subset of documents: =MyClass.update_documents(documents_hashes)= IMPORTANT: =documents_hashes= must be an array of hashes that
|
407
|
+
were generated via =search_indexable_hash= The =PRIMARY_SEARCH_KEY=
|
408
|
+
(=_id= by default) will be used to find records in the index to update.
|
409
|
+
|
410
|
+
*Deleting*
|
411
|
+
- Delete everything: =MyClass.delete_all_documents!=
|
412
|
+
- Delete a specific record: =my_instance.remove_from_search=
|
413
|
+
- Delete the index: =MyClass.delete_index!=
|
414
|
+
WARNING: if you think you should use this, /you're probably mistaken/.
|
415
|
+
|
416
|
+
**** Indexes
|
417
|
+
By default every model gets its own search index. This means that
|
418
|
+
=Foo.search("some text")= will only search =Foo= objects. To have a
|
419
|
+
search cross objects you'll need to use a "Shared Index" (see below).
|
420
|
+
|
421
|
+
The name of the index isn't important when not using shared indexes. By
|
422
|
+
default a model's index is the snake cased form of the class name. For
|
423
|
+
example, data for =MyWidget= models will be stored in the =my_widget=
|
424
|
+
index.
|
425
|
+
|
426
|
+
**** Shared indexes
|
427
|
+
Imagine you have a =Note= and a =Comment= model, sharing an index so
|
428
|
+
that you can perform a single search and have search results for both
|
429
|
+
models that are ranked by relevance.
|
430
|
+
|
431
|
+
In this case both models would define a =SEARCH_INDEX_NAME= constant
|
432
|
+
with the same value. You might want to just put this, and the other
|
433
|
+
search stuff in a common module that they all =include=.
|
434
|
+
|
435
|
+
Then, when you search you can say =Note.search("search term")= and it
|
436
|
+
will /only/ bring back results for =Note= records. If you want to
|
437
|
+
include results that match =Comment= records too, you can set the
|
438
|
+
optional =filtered_by_class= parameter to =false=.
|
439
|
+
|
440
|
+
For example: =Note.search("search term", filtered_by_class: false)= will
|
441
|
+
return all matching =Note= results, as well as results for /all/ the
|
442
|
+
other models that share the same index as =Note=.
|
443
|
+
|
444
|
+
⚠ Models sharing the same index must share the same primary key field as
|
445
|
+
well. This is a known limitation of the system.
|
446
|
+
|
447
|
+
** Searching
|
448
|
+
To get a list of all the matching objects in the order returned by the
|
449
|
+
search engine run =MyModel.search("search term")= Note that this will
|
450
|
+
restrict the results to records generated by the model you're calling
|
451
|
+
this on. If you have an index that contains data from multiple models
|
452
|
+
and wish to include all of them in the results pass in the optional
|
453
|
+
=filtered_by_class= parameter with a =false= value. E.g.
|
454
|
+
=MyModel.search("search term", filtered_by_class: false)=
|
455
|
+
|
456
|
+
Searching returns a hash, with the class name of the results as the key
|
457
|
+
and an array of String ids, or =Mongoid::Document= objects as the value.
|
458
|
+
By default it assumes you want =Mongoid::Document= objects. The returned
|
459
|
+
hash /also/ includes a key of ="search_result_metadata"= which includes
|
460
|
+
the metadata provided by Meilisearch regarding your request. You'll need
|
461
|
+
this for pagination if you have lots of results. To /exclude/ the
|
462
|
+
metadata pass =include_metadata: false= as an option. E.g.
|
463
|
+
=MyModel.search("search term", include_metadata: false)=
|
464
|
+
|
465
|
+
*** Useful Keyword Parameters
|
466
|
+
- =ids_only=
|
467
|
+
- only return matching ids. These will be an array under the
|
468
|
+
="matches"= key.
|
469
|
+
- defaults to =false=
|
470
|
+
- =filtered_by_class=
|
471
|
+
- limit results to the class you initiated the search from. E.g.
|
472
|
+
=Note.search("foo")= will only return results from the =Note= class
|
473
|
+
even if there are records from other classes in the same index.
|
474
|
+
- defaults to =true=
|
475
|
+
- =include_metadata=
|
476
|
+
- include the metadata about the search results provided by
|
477
|
+
Meilisearch. If true (default) there will be a
|
478
|
+
="search_result_metadata"= key, with a hash of the Meilisearch
|
479
|
+
metadata.
|
480
|
+
- You'll likely need this in order to support pagination, however if
|
481
|
+
you just want to return a single page worth of data, you can set
|
482
|
+
this to =false= to discard it.
|
483
|
+
- defaults to =true=
|
484
|
+
|
485
|
+
*** Example Search Results
|
486
|
+
Search results, ids only, for a class where
|
487
|
+
=CLASS_PREFIXED_SEARCH_IDS=false=.
|
488
|
+
|
489
|
+
#+begin_src ruby
|
490
|
+
Note.search('foo', ids_only: true) # => returns
|
491
|
+
{
|
492
|
+
"matches" => [
|
493
|
+
"64274a5d906b1d7d02c1fcc7",
|
494
|
+
"643f5e1c906b1d60f9763071",
|
495
|
+
"64483e63906b1d84f149717a"
|
496
|
+
],
|
497
|
+
"search_result_metadata" => {
|
498
|
+
"query"=>query_string,
|
499
|
+
"processingTimeMs"=>1,
|
500
|
+
"limit"=>50,
|
501
|
+
"offset"=>0,
|
502
|
+
"estimatedTotalHits"=>33,
|
503
|
+
"nbHits"=>33
|
504
|
+
}
|
505
|
+
}
|
506
|
+
#+end_src
|
507
|
+
|
508
|
+
If =CLASS_PREFIXED_SEARCH_IDS=true= the above would have ids like
|
509
|
+
="Note_64274a5d906b1d7d02c1fcc7"=
|
510
|
+
|
511
|
+
Without =ids_only= you get full objects in a =matches= array.
|
512
|
+
|
513
|
+
#+begin_src ruby
|
514
|
+
Note.search('foo') # or Note.search('foo', ids_only: false) # => returns
|
515
|
+
{
|
516
|
+
"matches" => [
|
517
|
+
#<Note _id: 64274a5d906b1d7d02c1fcc7, created_at: 2023-03-15 00:00:00 UTC, updated_at: 2023-03-31 21:02:21.108 UTC, title: "A note from the past", body: "a body", type: "misc", context: "dachary">,
|
518
|
+
#<Note _id: 643f5e1c906b1d60f9763071, created_at: 2023-04-18 00:00:00 UTC, updated_at: 2023-04-19 03:21:00.41 UTC, title: "offline standup ", body: "onother body", type: "misc", context: "WORK">,
|
519
|
+
#<Note _id: 64483e63906b1d84f149717a, created_at: 2023-04-25 00:00:00 UTC, updated_at: 2023-04-26 11:23:38.125 UTC, title: "Standup Notes (for wed)", body: "very full bodied", type: "misc", context: "WORK">
|
520
|
+
],
|
521
|
+
"search_result_metadata" => {
|
522
|
+
"query"=>query_string, "processingTimeMs"=>1, "limit"=>50,
|
523
|
+
"offset"=>0, "estimatedTotalHits"=>33, "nbHits"=>33
|
524
|
+
}
|
525
|
+
}
|
526
|
+
#+end_src
|
527
|
+
|
528
|
+
If =Note= records shared an index with =Task= and they both had
|
529
|
+
=CLASS_PREFIXED_SEARCH_ID=true= you'd get a result like this.
|
530
|
+
|
531
|
+
#+begin_src ruby
|
532
|
+
Note.search('foo') #=> returns
|
533
|
+
{
|
534
|
+
"matches" => [
|
535
|
+
#<Note _id: 64274a5d906b1d7d02c1fcc7, created_at: 2023-03-15 00:00:00 UTC, updated_at: 2023-03-31 21:02:21.108 UTC, title: "A note from the past", body: "a body", type: "misc", context: "dachary">,
|
536
|
+
#<Note _id: 643f5e1c906b1d60f9763071, created_at: 2023-04-18 00:00:00 UTC, updated_at: 2023-04-19 03:21:00.41 UTC, title: "offline standup ", body: "onother body", type: "misc", context: "WORK">,
|
537
|
+
#<Task _id: 64483e63906b1d84f149717a, created_at: 2023-04-25 00:00:00 UTC, updated_at: 2023-04-26 11:23:38.125 UTC, title: "Do the thing", body: "very full bodied", type: "misc", context: "WORK">
|
538
|
+
],
|
539
|
+
"search_result_metadata" => {
|
540
|
+
"query"=>query_string, "processingTimeMs"=>1, "limit"=>50,
|
541
|
+
"offset"=>0, "estimatedTotalHits"=>33, "nbHits"=>33
|
542
|
+
}
|
543
|
+
|
544
|
+
}
|
545
|
+
#+end_src
|
546
|
+
|
547
|
+
*** Custom Search Options
|
548
|
+
To invoke any of Meilisearch's custom search options (see
|
549
|
+
[[https://www.meilisearch.com/docs/reference/api/search][their
|
550
|
+
documentation]]). You can pass them in via an options hash.
|
551
|
+
|
552
|
+
=MyModel.search("search term", options: <my custom options>)=
|
553
|
+
|
554
|
+
Currently the Meilisearch-ruby gem can convert keys from snake case to
|
555
|
+
camel case. For example =hits_per_page= will become =hitsPerPage=.
|
556
|
+
Meilisearch ultimately wants camel case (=camelCase=) parameter keys,
|
557
|
+
/but/ =meilisearch-ruby= wants snake case (=snake_case=).
|
558
|
+
|
559
|
+
Follow Meilisearch's documentation to see what's available and what type
|
560
|
+
of options to pass it, but convert them to snake case first. Note that
|
561
|
+
your options keys and values must all be simple JSON values.
|
562
|
+
|
563
|
+
If for some reason that still isn't enough, you can work with the
|
564
|
+
meilisearch-ruby index directly via
|
565
|
+
=Search::Client.instance.index(search_index_name)=
|
566
|
+
|
567
|
+
**** Pagination
|
568
|
+
This gem has no specific pagination handling, as there are multiple
|
569
|
+
libraries for handling pagination in Ruby. Here's an example of how to
|
570
|
+
get started with [[https://github.com/ddnexus/pagy][Pagy]].
|
571
|
+
|
572
|
+
#+begin_src ruby
|
573
|
+
current_page_number = 1
|
574
|
+
max_items_per_page = 10
|
575
|
+
|
576
|
+
search_results = Note.search('foo')
|
577
|
+
|
578
|
+
Pagy.new(
|
579
|
+
count: search_results["search_result_metadata"]["nbHits"],
|
580
|
+
page: current_page_number,
|
581
|
+
items: max_items_per_page
|
582
|
+
)
|
583
|
+
#+end_src
|
584
|
+
|
585
|
+
** Upgrading
|
586
|
+
*** From v1.x
|
587
|
+
primary difference between v1.x and v2 is that =mongodb_meilisearch= no longer uses the Meilisearch master key [[https://www.meilisearch.com/docs/learn/security/basic_security][for security reasons]]. The /only/ thing it will use it for is to look up the default search API key and / or the default administrative API key.
|
588
|
+
|
589
|
+
It now initializes 2 =Meilisearch::Client= objects: one for searching, one for administration, and all of the internal methods that interact with an index have been rewritten to interact with it via the correct client.
|
590
|
+
|
591
|
+
|
592
|
+
Practically speaking, there are 3 significant differences from v1.x
|
593
|
+
|
594
|
+
1. =MEILISEARCH_API_KEY= is no longer supported. It has been replaced with the official Meilisearch environment variable for the same purpose: =MEILI_MASTER_KEY=
|
595
|
+
2. The master key is only used when an API key for search and / or administration is not provided. It is recommended that you provide =MEILISEARCH_SEARCH_KEY= & =MEILISEARCH_ADMIN_KEY= for those purposes. See "Finding your Search & Admin keys" above for more details.
|
596
|
+
3. =Search::Client.instance.client= has been replaced with 2 clients with specific purposes
|
597
|
+
- =Search::Client::instance.search_client=
|
598
|
+
- =Search::Client::instance.admin_client=
|
599
|
+
4. =ClassMethods.search_index= has been replaced with
|
600
|
+
- =ClassMethods.searchable_index=
|
601
|
+
- =ClassMethods.administratable_index=
|
602
|
+
|
603
|
+
** Development
|
604
|
+
To contribute to this gem.
|
605
|
+
|
606
|
+
- Run =bundle install= to install all the dependencies.
|
607
|
+
- run =lefthook install= to set up
|
608
|
+
[[https://github.com/evilmartians/lefthook][lefthook]] This will do
|
609
|
+
things like make sure the tests still pass, and run rubocop before you
|
610
|
+
commit.
|
611
|
+
- Start hacking.
|
612
|
+
- Add RSpec tests.
|
613
|
+
- Add your name to CONTRIBUTORS.md
|
614
|
+
- Make PR.
|
615
|
+
|
616
|
+
NOTE: by contributing to this repository you are offering to transfer
|
617
|
+
copyright to the current maintainer of the repository.
|
618
|
+
|
619
|
+
To install this gem onto your local machine, run
|
620
|
+
=bundle exec rake install=. To release a new version, update the version
|
621
|
+
number in =version.rb=, and then run =bundle exec rake release=, which
|
622
|
+
will create a git tag for the version, push git commits and the created
|
623
|
+
tag, and push the =.gem= file to [[https://rubygems.org][rubygems.org]].
|
624
|
+
|
625
|
+
Bug reports and pull requests are welcome on GitHub at
|
626
|
+
https://github.com/masukomi/mongodb_meilisearch. This project is
|
627
|
+
intended to be a safe, welcoming space for collaboration, and
|
628
|
+
contributors are expected to adhere to the
|
629
|
+
[[https://github.com/masukomi/mongodb_meilisearch/blob/main/CODE_OF_CONDUCT.md][code of conduct]].
|
630
|
+
|
631
|
+
** License
|
632
|
+
The gem is available as open source under the terms of the
|
633
|
+
[[https://github.com/masukomi/mongodb_meilisearch/blob/main/LICENSE.txt][Server Side Public License]]. For those unfamiliar, the short version is that
|
634
|
+
if you use it in a server side app you need to share all the code for
|
635
|
+
that app and its infrastructure. It's like AGPL on steroids. Commercial
|
636
|
+
licenses are available if you want to use this in a commercial setting
|
637
|
+
but not share all your source.
|
638
|
+
|
639
|
+
** Code of Conduct
|
640
|
+
Everyone interacting in this project's codebases, issue trackers, chat
|
641
|
+
rooms and mailing lists is expected to follow the [[https://github.com/masukomi/mongodb_meilisearch/blob/main/CODE_OF_CONDUCT.md][code of conduct]].
|
642
|
+
|
643
|
+
It is furthermore /fully/ acknowledged that "master" is a problematic term that shouldn't be used. However, we're stuck with it until Meilisearch changes the name of that key.
|
644
|
+
|
645
|
+
* Footnotes
|
646
|
+
[fn:macos] MacOS users I have not found a way to successfully automatically Meilisearch with launchctl (brew services uses this) that allows it to see the ~MEILI_MASTER_KEY~ environment variable. As such, I can /not/ recommend using launchctl for running Meilisearch. Running it without a master key only works in development mode, and introduces complications.
|
647
|
+
|
648
|
+
[fn:default_keys] Using the master key you can access the [[https://www.meilisearch.com/docs/reference/api/keys][/keys]] endpoint in Meilisearch to retrieve the default Search & Administrative API keys
|