1 ;;; registry.el --- Track and remember data items by various fields
3 ;; Copyright (C) 2011 Free Software Foundation, Inc.
5 ;; Author: Teodor Zlatanov <tzz@lifelogs.com>
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/>.
25 ;; This library provides a general-purpose EIEIO-based registry
26 ;; database with persistence, initialized with these fields:
28 ;; version: a float, 0.1 currently (don't change it)
30 ;; max-hard: an integer, default 5000000
32 ;; max-soft: an integer, default 50000
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. At
61 ;; PRUNE TIME (when the :prune-function is called), the registry will
62 ;; trim any entries without the F2 field until the size is :max-soft
63 ;; or less. No entries with the F2 field will be removed at PRUNE
66 ;; When an entry is inserted, the registry will reject new entries
67 ;; if they bring it over the max-hard limit, even if they have the F2
70 ;; The user decides which fields are "tracked", F1 for example. Any
71 ;; new entry is then indexed by all the tracked fields so it can be
72 ;; quickly looked up that way. The data is always a list (see example
73 ;; above) and each list element is indexed.
75 ;; Precious and tracked field names must be symbols. All other
76 ;; fields can be any other Emacs Lisp types.
80 (eval-when-compile (require 'cl))
83 (when (null (ignore-errors (require 'ert)))
84 (defmacro* ert-deftest (name () &body docstring-keys-and-body))))
89 (or (ignore-errors (progn
91 (require 'eieio-base)))
92 ;; gnus-fallback-lib/ from gnus/lisp/gnus-fallback-lib
94 (let ((load-path (cons (expand-file-name
95 "gnus-fallback-lib/eieio"
96 (file-name-directory (locate-library "gnus")))
99 (require 'eieio-base)))
101 "eieio not found in `load-path' or gnus-fallback-lib/ directory.")))
103 (defclass registry-db (eieio-persistent)
104 ((version :initarg :version
108 :documentation "The registry version.")
109 (max-hard :initarg :max-hard
113 :documentation "Never accept more than this many elements.")
114 (max-soft :initarg :max-soft
118 :documentation "Prune as much as possible to get to this size.")
119 (tracked :initarg :tracked
122 :documentation "The tracked (indexed) fields, a list of symbols.")
123 (precious :initarg :precious
126 :documentation "The precious fields, a list of symbols.")
127 (tracker :initarg :tracker
129 :documentation "The field tracking hashtable.")
132 :documentation "The data hashtable.")))
135 (defmethod initialize-instance :AFTER ((this registry-db) slots)
136 "Set value of data slot of THIS after initialization."
137 (with-slots (data tracker) this
138 (unless (member :data slots)
140 (make-hash-table :size 10000 :rehash-size 2.0 :test 'equal)))
141 (unless (member :tracker slots)
142 (setq tracker (make-hash-table :size 100 :rehash-size 2.0)))))
144 (defmethod registry-lookup ((db registry-db) keys)
145 "Search for KEYS in the registry-db THIS.
146 Returns a alist of the key followed by the entry in a list, not a cons cell."
147 (let ((data (oref db :data)))
151 (when (gethash k data)
152 (list k (gethash k data))))
155 (defmethod registry-lookup-breaks-before-lexbind ((db registry-db) keys)
156 "Search for KEYS in the registry-db THIS.
157 Returns a alist of the key followed by the entry in a list, not a cons cell."
158 (let ((data (oref db :data)))
160 (loop for key in keys
161 when (gethash key data)
162 collect (list key (gethash key data))))))
164 (defmethod registry-lookup-secondary ((db registry-db) tracksym
166 "Search for TRACKSYM in the registry-db THIS.
167 When CREATE is not nil, create the secondary index hashtable if needed."
168 (let ((h (gethash tracksym (oref db :tracker))))
173 (make-hash-table :size 800 :rehash-size 2.0 :test 'equal)
175 (gethash tracksym (oref db :tracker))))))
177 (defmethod registry-lookup-secondary-value ((db registry-db) tracksym val
179 "Search for TRACKSYM with value VAL in the registry-db THIS.
180 When SET is not nil, set it for VAL (use t for an empty list)."
181 ;; either we're asked for creation or there should be an existing index
182 (when (or set (registry-lookup-secondary db tracksym))
183 ;; set the entry if requested,
185 (puthash val (if (eq t set) '() set)
186 (registry-lookup-secondary db tracksym t)))
187 (gethash val (registry-lookup-secondary db tracksym)))))
189 (defun registry--match (mode entry check-list)
192 (let ((key (nth 0 (nth 0 check-list)))
193 (vals (cdr-safe (nth 0 check-list)))
195 (while (and key vals (not found))
196 (setq found (case mode
198 (member (car-safe vals) (cdr-safe (assoc key entry))))
200 (string-match (car vals)
203 (cdr-safe (assoc key entry))
205 vals (cdr-safe vals)))
207 (registry--match mode entry (cdr-safe check-list))))))
210 (defmethod registry-search ((db registry-db) &rest spec)
211 "Search for SPEC across the registry-db THIS.
212 For example calling with :member '(a 1 2) will match entry '((a 3 1)).
213 Calling with :all t (any non-nil value) will match all.
214 Calling with :regex '\(a \"h.llo\") will match entry '((a \"hullo\" \"bye\").
215 The test order is to check :all first, then :member, then :regex."
217 (let ((all (plist-get spec :all))
218 (member (plist-get spec :member))
219 (regex (plist-get spec :regex)))
220 (loop for k being the hash-keys of (oref db :data)
221 using (hash-values v)
223 ;; :all non-nil returns all
226 (and member (registry--match :member v member))
228 (and regex (registry--match :regex v regex)))
231 (defmethod registry-delete ((db registry-db) keys assert &rest spec)
232 "Delete KEYS from the registry-db THIS.
233 If KEYS is nil, use SPEC to do a search.
234 Updates the secondary ('tracked') indices as well.
235 With assert non-nil, errors out if the key does not exist already."
236 (let* ((data (oref db :data))
238 (apply 'registry-search db spec)))
239 (tracked (oref db :tracked)))
242 (let ((entry (gethash key data)))
245 "Key %s does not exists in database" key))
246 ;; clean entry from the secondary indices
248 ;; is this tracked symbol indexed?
249 (when (registry-lookup-secondary db tr)
250 ;; for every value in the entry under that key...
251 (dolist (val (cdr-safe (assq tr entry)))
252 (let* ((value-keys (registry-lookup-secondary-value
254 (when (member key value-keys)
255 ;; override the previous value
256 (registry-lookup-secondary-value
258 ;; with the indexed keys MINUS the current key
259 ;; (we pass t when the list is empty)
260 (or (delete key value-keys) t)))))))
264 (defmethod registry-insert ((db registry-db) key entry)
265 "Insert ENTRY under KEY into the registry-db THIS.
266 Updates the secondary ('tracked') indices as well.
267 Errors out if the key exists already."
269 (assert (not (gethash key (oref db :data))) nil
270 "Key already exists in database")
272 (assert (< (registry-size db)
275 "registry max-hard size limit reached")
278 (puthash key entry (oref db :data))
280 ;; store the secondary indices
281 (dolist (tr (oref db :tracked))
282 ;; for every value in the entry under that key...
283 (dolist (val (cdr-safe (assq tr entry)))
284 (let* ((value-keys (registry-lookup-secondary-value db tr val)))
285 (pushnew key value-keys :test 'equal)
286 (registry-lookup-secondary-value db tr val value-keys))))
289 (defmethod registry-reindex ((db registry-db))
290 "Rebuild the secondary indices of registry-db THIS."
292 (expected (* (length (oref db :tracked)) (registry-size db))))
293 (dolist (tr (oref db :tracked))
298 (when (and (< 0 expected)
299 (= 0 (mod count 1000)))
300 (message "reindexing: %d of %d (%.2f%%)"
301 count expected (/ (* 100 count) expected)))
302 (dolist (val (cdr-safe (assq tr v)))
303 (let* ((value-keys (registry-lookup-secondary-value db tr val)))
304 (push key value-keys)
305 (registry-lookup-secondary-value db tr val value-keys))))
308 (defmethod registry-size ((db registry-db))
309 "Returns the size of the registry-db object THIS.
310 This is the key count of the :data slot."
311 (hash-table-count (oref db :data)))
313 (defmethod registry-prune ((db registry-db))
314 "Prunes the registry-db object THIS.
315 Removes only entries without the :precious keys."
316 (let* ((precious (oref db :precious))
317 (precious-p (lambda (entry-key)
318 (cdr (memq (car entry-key) precious))))
319 (data (oref db :data))
320 (limit (oref db :max-soft))
321 (size (registry-size db))
322 (candidates (loop for k being the hash-keys of data
323 using (hash-values v)
324 when (notany precious-p v)
326 (candidates-count (length candidates))
327 ;; are we over max-soft?
328 (prune-needed (> size limit)))
330 ;; while we have more candidates than we need to remove...
331 (while (and (> candidates-count (- size limit)) candidates)
332 (decf candidates-count)
333 (setq candidates (cdr candidates)))
335 (registry-delete db candidates nil))))
337 (ert-deftest registry-instantiation-test ()
338 (should (registry-db "Testing")))
340 (ert-deftest registry-match-test ()
341 (let ((entry '((hello "goodbye" "bye") (blank))))
343 (message "Testing :regex matching")
344 (should (registry--match :regex entry '((hello "nye" "bye"))))
345 (should (registry--match :regex entry '((hello "good"))))
346 (should-not (registry--match :regex entry '((hello "nye"))))
347 (should-not (registry--match :regex entry '((hello))))
349 (message "Testing :member matching")
350 (should (registry--match :member entry '((hello "bye"))))
351 (should (registry--match :member entry '((hello "goodbye"))))
352 (should-not (registry--match :member entry '((hello "good"))))
353 (should-not (registry--match :member entry '((hello "nye"))))
354 (should-not (registry--match :member entry '((hello)))))
355 (message "Done with matching testing."))
357 (defun registry-make-testable-db (n &optional name file)
358 (let* ((db (registry-db
360 :file (or file "unused")
362 :max-soft 0 ; keep nothing not precious
363 :precious '(extra more-extra)
364 :tracked '(sender subject groups))))
366 (registry-insert db i `((sender "me")
367 (subject "about you")
368 (more-extra) ; empty data key should be pruned
369 ;; first 5 entries will NOT have this extra data
370 ,@(when (< 5 i) (list (list 'extra "more data")))
371 (groups ,(number-to-string i)))))
374 (ert-deftest registry-usage-test ()
376 (db (registry-make-testable-db n)))
377 (message "size %d" n)
378 (should (= n (registry-size db)))
379 (message "max-hard test")
380 (should-error (registry-insert db "new" '()))
381 (message "Individual lookup")
382 (should (= 58 (caadr (registry-lookup db '(1 58 99)))))
383 (message "Grouped individual lookup")
384 (should (= 3 (length (registry-lookup db '(1 58 99)))))
385 (when (boundp 'lexical-binding)
386 (message "Individual lookup (breaks before lexbind)")
388 (caadr (registry-lookup-breaks-before-lexbind db '(1 58 99)))))
389 (message "Grouped individual lookup (breaks before lexbind)")
391 (length (registry-lookup-breaks-before-lexbind db
394 (should (= n (length (registry-search db :all t))))
395 (should (= n (length (registry-search db :member '((sender "me"))))))
396 (message "Secondary index search")
397 (should (= n (length (registry-lookup-secondary-value db 'sender "me"))))
398 (should (equal '(74) (registry-lookup-secondary-value db 'groups "74")))
400 (should (registry-delete db '(1) t))
402 (message "Search after delete")
403 (should (= n (length (registry-search db :all t))))
404 (message "Secondary search after delete")
405 (should (= n (length (registry-lookup-secondary-value db 'sender "me"))))
407 (let* ((tokeep (registry-search db :member '((extra "more data"))))
408 (count (- n (length tokeep)))
409 (pruned (registry-prune db))
410 (prune-count (length pruned)))
411 (message "Expecting to prune %d entries and pruned %d"
413 (should (and (= count 5)
414 (= count prune-count))))
415 (message "Done with usage testing.")))
417 (ert-deftest registry-persistence-test ()
419 (tempfile (make-temp-file "registry-persistence-"))
420 (name "persistence tester")
421 (db (registry-make-testable-db n name tempfile))
423 (message "Saving to %s" tempfile)
424 (eieio-persistent-save db)
425 (setq size (nth 7 (file-attributes tempfile)))
426 (message "Saved to %s: size %d" tempfile size)
429 (insert-file-contents-literally tempfile)
430 (should (looking-at (concat ";; Object "
432 "\n;; EIEIO PERSISTENT OBJECT"))))
433 (message "Reading object back")
434 (setq back (eieio-persistent-read tempfile))
436 (message "Read object back: %d keys, expected %d==%d"
437 (registry-size back) n (registry-size db))
438 (should (= (registry-size back) n))
439 (should (= (registry-size back) (registry-size db)))
440 (delete-file tempfile))
441 (message "Done with persistence testing."))
444 ;;; registry.el ends here