[elpa] externals/ebdb a43e883 026/350: Move record initialization out of load process

[Eric Abrahamsen]

branch: externals/ebdb
commit a43e883eaa3f8c5360394952717ef2c598bf99cf
Author: Eric Abrahamsen <address@hidden>
Commit: Eric Abrahamsen <address@hidden>

    Move record initialization out of load process
    
    Part of the motivation of this is that initialization creates
    connections between records.  If those records are living in separate
    databases, then all databases need to be loaded before initialization
    happens.
    
    * ebdb.el (ebdb-db-load): Don't initialize records here.
      (ebdb-initialize): Instead, create a separate function for doing
      that.
      (ebdb-load): And call it here, after all databases are loaded.
---
 ebdb.el | 40 +++++++++++++++++++---------------------
 1 file changed, 19 insertions(+), 21 deletions(-)

diff --git a/ebdb.el b/ebdb.el
index 94a753a..5387ec8 100644
--- a/ebdb.el
+++ b/ebdb.el
@@ -1736,17 +1736,12 @@ from its respective databases, un-hashing its uuid, 
running
       (cl-call-next-method record slots))))
 
 (cl-defmethod ebdb-init ((record ebdb-record))
-  "Initiate a record after loading a database or creating a new record.
-
-Call `ebdb-init' on all of a record's user fields."
+  "Initiate a record after loading a database or creating a new
+record."
   (dolist (field (ebdb-record-user-fields record))
     (ebdb-init field record))
   (let ((uuid (ebdb-record-uuid record)))
-    (ebdb-puthash uuid record)
-    ;; I don't think there are any cases when we would be at this
-    ;; point in the code and the uuid would already be seen, but...
-    (unless (member uuid ebdb-seen-uuids)
-      (push uuid ebdb-seen-uuids)))
+    (ebdb-puthash uuid record))
   (ebdb-record-set-sortkey record))
 
 (cl-defmethod ebdb-merge ((left ebdb-record)
@@ -2781,17 +2776,12 @@ overwrite data somewhere."
     ;; Cycle over each loaded record.
     (condition-case err
        (progn
-         ;; First, tell it about the database.
+         ;; Tell it about the database.
          (object-add-to-list (ebdb-record-cache rec)
                              'database db)
 
-         ;; Make sure its UUID is unique.
-         (ebdb-check-uuid (ebdb-record-uuid rec))
-
-         ;; Lastly initialize it.  This has to come after
-         ;; `ebdb-check-uuid' because it adds the record's UUID to
-         ;; `ebdb-seen-uuids'.
-         (ebdb-init rec))
+         ;; Make sure its UUID is unique.  Doesn't create new UUIDs.
+         (ebdb-check-uuid (ebdb-record-uuid rec)))
 
       (ebdb-duplicate-uuid
        ;; There's a duplicate, decide what to do about it.  In the
@@ -2837,11 +2827,7 @@ overwrite data somewhere."
           (object-add-to-list db 'records keeper)
           (object-add-to-list (ebdb-record-cache keeper)
                               'database d))
-        (ebdb-delete deleter)
-        ;; Double was initialized, rec wasn't.  If we end up keeping
-        ;; rec, we need to initialize it.
-        (when delete-double
-          (ebdb-init keeper)))))))
+        (ebdb-delete deleter))))))
 
 (cl-defmethod ebdb-db-unload ((db ebdb-db))
   "Unload database DB.
@@ -4320,11 +4306,23 @@ important work is done by the `ebdb-db-load' method."
        ;; We're migrating from a previous version of EBDB.
        (ebdb-migrate-from-bbdb)
       (message "Loading EBDB sources... done"))
+    (ebdb-initialize)
     ;; Users will expect the same ordering as `ebdb-sources'
     (setq ebdb-db-list (nreverse ebdb-db-list))
     (run-hooks 'ebdb-after-load-hook)
     (length ebdb-record-tracker)))
 
+;; If we wanted to make it seem like EBDB was loading faster, we could
+;; defer calls to `ebdb-initialize' until the first time the database
+;; is searched.
+(defun ebdb-initialize ()
+  "After all databases are loaded, initialize the records.
+
+This results in the creation of all the secondary data
+structures: label lists, `ebdb-org-hashtable', record caches,
+etc."
+  (mapcar #'ebdb-init ebdb-record-tracker))
+
 (defun ebdb-address-continental-p (address)
   "Return non-nil if ADDRESS is a continental address.
 This is done by comparing the postcode to `ebdb-continental-postcode-regexp'.