* lisp/mail/rmail.el: Use lexical-binding
[emacs.git] / lisp / gnus / registry.el
blob96a89fca1b707cbcc9ce938e3657bec1b1ffee2b
1 ;;; registry.el --- Track and remember data items by various fields
3 ;; Copyright (C) 2011-2015 Free Software Foundation, Inc.
5 ;; Author: Teodor Zlatanov <tzz@lifelogs.com>
6 ;; Keywords: data
8 ;; This file is part of GNU Emacs.
10 ;; GNU Emacs is free software: you can redistribute it and/or modify
11 ;; it under the terms of the GNU General Public License as published by
12 ;; the Free Software Foundation, either version 3 of the License, or
13 ;; (at your option) any later version.
15 ;; GNU Emacs is distributed in the hope that it will be useful,
16 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
17 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 ;; GNU General Public License for more details.
20 ;; You should have received a copy of the GNU General Public License
21 ;; along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>.
23 ;;; Commentary:
25 ;; This library provides a general-purpose EIEIO-based registry
26 ;; database with persistence, initialized with these fields:
28 ;; version: a float
30 ;; max-size: an integer, default most-positive-fixnum
32 ;; prune-factor: a float between 0 and 1, default 0.1
34 ;; precious: a list of symbols
36 ;; tracked: a list of symbols
38 ;; tracker: a hashtable tuned for 100 symbols to track (you should
39 ;; only access this with the :lookup2-function and the
40 ;; :lookup2+-function)
42 ;; data: a hashtable with default size 10K and resize threshold 2.0
43 ;; (this reflects the expected usage so override it if you know better)
45 ;; ...plus methods to do all the work: `registry-search',
46 ;; `registry-lookup', `registry-lookup-secondary',
47 ;; `registry-lookup-secondary-value', `registry-insert',
48 ;; `registry-delete', `registry-prune', `registry-size' which see
50 ;; and with the following properties:
52 ;; Every piece of data has a unique ID and some general-purpose fields
53 ;; (F1=D1, F2=D2, F3=(a b c)...) expressed as an alist, e.g.
55 ;; ((F1 D1) (F2 D2) (F3 a b c))
57 ;; Note that whether a field has one or many pieces of data, the data
58 ;; is always a list of values.
60 ;; The user decides which fields are "precious", F2 for example. When
61 ;; the registry is pruned, any entries without the F2 field will be
62 ;; removed until the size is :max-size * :prune-factor _less_ than the
63 ;; maximum database size. No entries with the F2 field will be removed
64 ;; at PRUNE TIME, which means it may not be possible to prune back all
65 ;; the way to the target size.
67 ;; When an entry is inserted, the registry will reject new entries if
68 ;; they bring it over the :max-size limit, even if they have the F2
69 ;; field.
71 ;; The user decides which fields are "tracked", F1 for example. Any
72 ;; new entry is then indexed by all the tracked fields so it can be
73 ;; quickly looked up that way. The data is always a list (see example
74 ;; above) and each list element is indexed.
76 ;; Precious and tracked field names must be symbols. All other
77 ;; fields can be any other Emacs Lisp types.
79 ;;; Code:
81 (eval-when-compile (require 'cl))
83 (require 'eieio)
84 (require 'eieio-base)
86 ;; The version number needs to be kept outside of the class definition
87 ;; itself. The persistent-save process does *not* write to file any
88 ;; slot values that are equal to the default :initform value. If a
89 ;; database object is at the most recent version, therefore, its
90 ;; version number will not be written to file. That makes it
91 ;; difficult to know when a database needs to be upgraded.
92 (defvar registry-db-version 0.2
93 "The current version of the registry format.")
95 (defclass registry-db (eieio-persistent)
96 ((version :initarg :version
97 :initform nil
98 :type (or null float)
99 :documentation "The registry version.")
100 (max-size :initarg :max-size
101 ;; EIEIO's :initform is not 100% compatible with CLOS in
102 ;; that if the form is an atom, it assumes it's constant
103 ;; value rather than an expression, so in order to get the value
104 ;; of `most-positive-fixnum', we need to use an
105 ;; expression that's not just a symbol.
106 :initform (symbol-value 'most-positive-fixnum)
107 :type integer
108 :custom integer
109 :documentation "The maximum number of registry entries.")
110 (prune-factor
111 :initarg :prune-factor
112 :initform 0.1
113 :type float
114 :custom float
115 :documentation "Prune to \(:max-size * :prune-factor\) less
116 than the :max-size limit. Should be a float between 0 and 1.")
117 (tracked :initarg :tracked
118 :initform nil
119 :type t
120 :documentation "The tracked (indexed) fields, a list of symbols.")
121 (precious :initarg :precious
122 :initform nil
123 :type t
124 :documentation "The precious fields, a list of symbols.")
125 (tracker :initarg :tracker
126 :type hash-table
127 :documentation "The field tracking hashtable.")
128 (data :initarg :data
129 :type hash-table
130 :documentation "The data hashtable.")))
132 (defmethod initialize-instance :BEFORE ((this registry-db) slots)
133 "Check whether a registry object needs to be upgraded."
134 ;; Hardcoded upgrade routines. Version 0.1 to 0.2 requires the
135 ;; :max-soft slot to disappear, and the :max-hard slot to be renamed
136 ;; :max-size.
137 (let ((current-version
138 (and (plist-member slots :version)
139 (plist-get slots :version))))
140 (when (or (null current-version)
141 (eql current-version 0.1))
142 (setq slots
143 (plist-put slots :max-size (plist-get slots :max-hard)))
144 (setq slots
145 (plist-put slots :version registry-db-version))
146 (cl-remf slots :max-hard)
147 (cl-remf slots :max-soft))))
149 (defmethod initialize-instance :AFTER ((this registry-db) slots)
150 "Set value of data slot of THIS after initialization."
151 (with-slots (data tracker) this
152 (unless (member :data slots)
153 (setq data
154 (make-hash-table :size 10000 :rehash-size 2.0 :test 'equal)))
155 (unless (member :tracker slots)
156 (setq tracker (make-hash-table :size 100 :rehash-size 2.0)))))
158 (defmethod registry-lookup ((db registry-db) keys)
159 "Search for KEYS in the registry-db THIS.
160 Returns an alist of the key followed by the entry in a list, not a cons cell."
161 (let ((data (oref db data)))
162 (delq nil
163 (mapcar
164 (lambda (k)
165 (when (gethash k data)
166 (list k (gethash k data))))
167 keys))))
169 (defmethod registry-lookup-breaks-before-lexbind ((db registry-db) keys)
170 "Search for KEYS in the registry-db THIS.
171 Returns an alist of the key followed by the entry in a list, not a cons cell."
172 (let ((data (oref db data)))
173 (delq nil
174 (loop for key in keys
175 when (gethash key data)
176 collect (list key (gethash key data))))))
178 (defmethod registry-lookup-secondary ((db registry-db) tracksym
179 &optional create)
180 "Search for TRACKSYM in the registry-db THIS.
181 When CREATE is not nil, create the secondary index hashtable if needed."
182 (let ((h (gethash tracksym (oref db :tracker))))
183 (if h
185 (when create
186 (puthash tracksym
187 (make-hash-table :size 800 :rehash-size 2.0 :test 'equal)
188 (oref db tracker))
189 (gethash tracksym (oref db tracker))))))
191 (defmethod registry-lookup-secondary-value ((db registry-db) tracksym val
192 &optional set)
193 "Search for TRACKSYM with value VAL in the registry-db THIS.
194 When SET is not nil, set it for VAL (use t for an empty list)."
195 ;; either we're asked for creation or there should be an existing index
196 (when (or set (registry-lookup-secondary db tracksym))
197 ;; set the entry if requested,
198 (when set
199 (puthash val (if (eq t set) '() set)
200 (registry-lookup-secondary db tracksym t)))
201 (gethash val (registry-lookup-secondary db tracksym))))
203 (defun registry--match (mode entry check-list)
204 ;; for all members
205 (when check-list
206 (let ((key (nth 0 (nth 0 check-list)))
207 (vals (cdr-safe (nth 0 check-list)))
208 found)
209 (while (and key vals (not found))
210 (setq found (case mode
211 (:member
212 (member (car-safe vals) (cdr-safe (assoc key entry))))
213 (:regex
214 (string-match (car vals)
215 (mapconcat
216 'prin1-to-string
217 (cdr-safe (assoc key entry))
218 "\0"))))
219 vals (cdr-safe vals)))
220 (or found
221 (registry--match mode entry (cdr-safe check-list))))))
223 (defmethod registry-search ((db registry-db) &rest spec)
224 "Search for SPEC across the registry-db THIS.
225 For example calling with :member '(a 1 2) will match entry '((a 3 1)).
226 Calling with :all t (any non-nil value) will match all.
227 Calling with :regex '\(a \"h.llo\") will match entry '((a \"hullo\" \"bye\").
228 The test order is to check :all first, then :member, then :regex."
229 (when db
230 (let ((all (plist-get spec :all))
231 (member (plist-get spec :member))
232 (regex (plist-get spec :regex)))
233 (loop for k being the hash-keys of (oref db data)
234 using (hash-values v)
235 when (or
236 ;; :all non-nil returns all
237 all
238 ;; member matching
239 (and member (registry--match :member v member))
240 ;; regex matching
241 (and regex (registry--match :regex v regex)))
242 collect k))))
244 (defmethod registry-delete ((db registry-db) keys assert &rest spec)
245 "Delete KEYS from the registry-db THIS.
246 If KEYS is nil, use SPEC to do a search.
247 Updates the secondary ('tracked') indices as well.
248 With assert non-nil, errors out if the key does not exist already."
249 (let* ((data (oref db data))
250 (keys (or keys
251 (apply 'registry-search db spec)))
252 (tracked (oref db tracked)))
254 (dolist (key keys)
255 (let ((entry (gethash key data)))
256 (when assert
257 (assert entry nil
258 "Key %s does not exist in database" key))
259 ;; clean entry from the secondary indices
260 (dolist (tr tracked)
261 ;; is this tracked symbol indexed?
262 (when (registry-lookup-secondary db tr)
263 ;; for every value in the entry under that key...
264 (dolist (val (cdr-safe (assq tr entry)))
265 (let* ((value-keys (registry-lookup-secondary-value
266 db tr val)))
267 (when (member key value-keys)
268 ;; override the previous value
269 (registry-lookup-secondary-value
270 db tr val
271 ;; with the indexed keys MINUS the current key
272 ;; (we pass t when the list is empty)
273 (or (delete key value-keys) t)))))))
274 (remhash key data)))
275 keys))
277 (defmethod registry-size ((db registry-db))
278 "Returns the size of the registry-db object THIS.
279 This is the key count of the `data' slot."
280 (hash-table-count (oref db data)))
282 (defmethod registry-full ((db registry-db))
283 "Checks if registry-db THIS is full."
284 (>= (registry-size db)
285 (oref db max-size)))
287 (defmethod registry-insert ((db registry-db) key entry)
288 "Insert ENTRY under KEY into the registry-db THIS.
289 Updates the secondary ('tracked') indices as well.
290 Errors out if the key exists already."
292 (assert (not (gethash key (oref db data))) nil
293 "Key already exists in database")
295 (assert (not (registry-full db))
296 nil
297 "registry max-size limit reached")
299 ;; store the entry
300 (puthash key entry (oref db data))
302 ;; store the secondary indices
303 (dolist (tr (oref db tracked))
304 ;; for every value in the entry under that key...
305 (dolist (val (cdr-safe (assq tr entry)))
306 (let* ((value-keys (registry-lookup-secondary-value db tr val)))
307 (pushnew key value-keys :test 'equal)
308 (registry-lookup-secondary-value db tr val value-keys))))
309 entry)
311 (defmethod registry-reindex ((db registry-db))
312 "Rebuild the secondary indices of registry-db THIS."
313 (let ((count 0)
314 (expected (* (length (oref db tracked)) (registry-size db))))
315 (dolist (tr (oref db tracked))
316 (let (values)
317 (maphash
318 (lambda (key v)
319 (incf count)
320 (when (and (< 0 expected)
321 (= 0 (mod count 1000)))
322 (message "reindexing: %d of %d (%.2f%%)"
323 count expected (/ (* 100 count) expected)))
324 (dolist (val (cdr-safe (assq tr v)))
325 (let* ((value-keys (registry-lookup-secondary-value db tr val)))
326 (push key value-keys)
327 (registry-lookup-secondary-value db tr val value-keys))))
328 (oref db data))))))
330 (defmethod registry-prune ((db registry-db) &optional sortfunc)
331 "Prunes the registry-db object DB.
333 Attempts to prune the number of entries down to \(*
334 :max-size :prune-factor\) less than the max-size limit, so
335 pruning doesn't need to happen on every save. Removes only
336 entries without the :precious keys, so it may not be possible to
337 reach the target limit.
339 Entries to be pruned are first sorted using SORTFUNC. Entries
340 from the front of the list are deleted first.
342 Returns the number of deleted entries."
343 (let ((size (registry-size db))
344 (target-size
345 (floor (- (oref db max-size)
346 (* (oref db max-size)
347 (oref db prune-factor)))))
348 candidates)
349 (if (registry-full db)
350 (progn
351 (setq candidates
352 (registry-collect-prune-candidates
353 db (- size target-size) sortfunc))
354 (length (registry-delete db candidates nil)))
355 0)))
357 (defmethod registry-collect-prune-candidates ((db registry-db) limit sortfunc)
358 "Collects pruning candidates from the registry-db object DB.
360 Proposes only entries without the :precious keys, and attempts to
361 return LIMIT such candidates. If SORTFUNC is provided, sort
362 entries first and return candidates from beginning of list."
363 (let* ((precious (oref db precious))
364 (precious-p (lambda (entry-key)
365 (cdr (memq (car entry-key) precious))))
366 (data (oref db data))
367 (candidates (cl-loop for k being the hash-keys of data
368 using (hash-values v)
369 when (notany precious-p v)
370 collect (cons k v))))
371 ;; We want the full entries for sorting, but should only return a
372 ;; list of entry keys.
373 (when sortfunc
374 (setq candidates (sort candidates sortfunc)))
375 (cl-subseq (mapcar #'car candidates) 0 (min limit (length candidates)))))
377 (provide 'registry)
378 ;;; registry.el ends here