emacs: Switch from text to JSON format for search results

The JSON format eliminates the complex escaping issues that have
plagued the text search format.  This uses the incremental JSON parser
so that, like the text parser, it can output search results
incrementally.

This slows down the parser by about ~4X, but puts us in a good
position to optimize either by improving the JSON parser (evidence
suggests this can reduce the overhead to ~40% over the text format) or
by switching to S-expressions (evidence suggests this will more than
double performance over the text parser).  [1]

This also fixes the incremental search parsing test.

This has one minor side-effect on search result formatting.
Previously, the date field was always padded to a fixed width of 12
characters because of how the text parser's regexp was written.  The
JSON format doesn't do this.  We could pad it out in Emacs before
formatting it, but, since all of the other fields are variable width,
we instead fix notmuch-search-result-format to take the variable-width
field and pad it out.  For users who have customized this variable,
we'll mention in the NEWS how to fix this slight format change.

[1] id:"20110720205007.GB21316@mit.edu"

emacs/notmuch.el

@@ -60,7 +60,7 @@
 (require 'notmuch-message)
 
 (defcustom notmuch-search-result-format
-  `(("date" . "%s ")
+  `(("date" . "%12s ")
     ("count" . "%-7s ")
     ("authors" . "%-20s ")
     ("subject" . "%s ")
@@ -557,17 +557,14 @@ This function advances the next thread when finished."
   (notmuch-search-tag '("-inbox"))
   (notmuch-search-next-thread))
 
-(defvar notmuch-search-process-filter-data nil
-  "Data that has not yet been processed.")
-(make-variable-buffer-local 'notmuch-search-process-filter-data)
-
 (defun notmuch-search-process-sentinel (proc msg)
   "Add a message to let user know when \"notmuch search\" exits"
   (let ((buffer (process-buffer proc))
 	(status (process-status proc))
 	(exit-status (process-exit-status proc))
 	(never-found-target-thread nil))
-    (if (memq status '(exit signal))
+    (when (memq status '(exit signal))
+	(kill-buffer (process-get proc 'parse-buf))
 	(if (buffer-live-p buffer)
 	    (with-current-buffer buffer
 	      (save-excursion
@@ -577,8 +574,6 @@ This function advances the next thread when finished."
 		  (if (eq status 'signal)
 		      (insert "Incomplete search results (search process was killed).\n"))
 		  (when (eq status 'exit)
-		    (if notmuch-search-process-filter-data
-			(insert (concat "Error: Unexpected output from notmuch search:\n" notmuch-search-process-filter-data)))
 		    (insert "End of search results.")
 		    (unless (= exit-status 0)
 		      (insert (format " (process returned %d)" exit-status)))
@@ -758,45 +753,59 @@ non-authors is found, assume that all of the authors match."
     (insert (apply #'format string objects))
     (insert "\n")))
 
+(defvar notmuch-search-process-state nil
+  "Parsing state of the search process filter.")
+
+(defvar notmuch-search-json-parser nil
+  "Incremental JSON parser for the search process filter.")
+
 (defun notmuch-search-process-filter (proc string)
   "Process and filter the output of \"notmuch search\""
-  (let ((buffer (process-buffer proc)))
+  (let ((results-buf (process-buffer proc))
-    (if (buffer-live-p buffer)
+	(parse-buf (process-get proc 'parse-buf))
-	(with-current-buffer buffer
+	(inhibit-read-only t)
-	    (let ((line 0)
+	done)
-		  (more t)
+    (if (not (buffer-live-p results-buf))
-		  (inhibit-read-only t)
+	(delete-process proc)
-		  (string (concat notmuch-search-process-filter-data string)))
+      (with-current-buffer parse-buf
-	      (setq notmuch-search-process-filter-data nil)
+	;; Insert new data
-	      (while more
+	(save-excursion
-		(while (and (< line (length string)) (= (elt string line) ?\n))
+	  (goto-char (point-max))
-		  (setq line (1+ line)))
+	  (insert string)))
-		(if (string-match "^thread:\\([0-9A-Fa-f]*\\) \\([^][]*\\) \\[\\([0-9]*\\)/\\([0-9]*\\)\\] \\([^;]*\\); \\(.*\\) (\\([^()]*\\))$" string line)
+      (with-current-buffer results-buf
-		    (let* ((thread-id (match-string 1 string))
+	(while (not done)
-			   (tags-str (match-string 7 string))
+	  (condition-case nil
-			   (result (list :thread thread-id
+	      (case notmuch-search-process-state
-					 :date_relative (match-string 2 string)
+		((begin)
-					 :matched (string-to-number
+		 ;; Enter the results list
-						   (match-string 3 string))
+		 (if (eq (notmuch-json-begin-compound
-					 :total (string-to-number
+			  notmuch-search-json-parser) 'retry)
-						 (match-string 4 string))
+		     (setq done t)
-					 :authors (match-string 5 string)
+		   (setq notmuch-search-process-state 'result)))
-					 :subject (match-string 6 string)
+		((result)
-					 :tags (if tags-str
+		 ;; Parse a result
-						   (save-match-data
+		 (let ((result (notmuch-json-read notmuch-search-json-parser)))
-						     (split-string tags-str))))))
+		   (case result
-		      (if (/= (match-beginning 0) line)
+		     ((retry) (setq done t))
-			  (notmuch-search-show-error
+		     ((end) (setq notmuch-search-process-state 'end))
-			   (substring string line (match-beginning 0))))
+		     (otherwise (notmuch-search-show-result result)))))
-		      (notmuch-search-show-result result)
+		((end)
-		      (set 'line (match-end 0)))
+		 ;; Any trailing data is unexpected
-		  (set 'more nil)
+		 (notmuch-json-eof notmuch-search-json-parser)
-		  (while (and (< line (length string)) (= (elt string line) ?\n))
+		 (setq done t)))
-		    (setq line (1+ line)))
+	    (json-error
-		  (if (< line (length string))
+	     ;; Do our best to resynchronize and ensure forward
-		      (setq notmuch-search-process-filter-data (substring string line)))
+	     ;; progress
-		  ))))
+	     (notmuch-search-show-error
-      (delete-process proc))))
+	      "%s"
+	      (with-current-buffer parse-buf
+		(let ((bad (buffer-substring (line-beginning-position)
+					     (line-end-position))))
+		  (forward-line)
+		  bad))))))
+	;; Clear out what we've parsed
+	(with-current-buffer parse-buf
+	  (delete-region (point-min) (point)))))))
 
 (defun notmuch-search-tag-all (&optional tag-changes)
   "Add/remove tags from all messages in current search buffer.
@@ -899,10 +908,19 @@ Other optional parameters are used as follows:
 	(let ((proc (start-process
 		     "notmuch-search" buffer
 		     notmuch-command "search"
+		     "--format=json"
 		     (if oldest-first
 			 "--sort=oldest-first"
 		       "--sort=newest-first")
-		     query)))
+		     query))
+	      ;; Use a scratch buffer to accumulate partial output.
+	      ;; This buffer will be killed by the sentinel, which
+	      ;; should be called no matter how the process dies.
+	      (parse-buf (generate-new-buffer " *notmuch search parse*")))
+	  (set (make-local-variable 'notmuch-search-process-state) 'begin)
+	  (set (make-local-variable 'notmuch-search-json-parser)
+	       (notmuch-json-create-parser parse-buf))
+	  (process-put proc 'parse-buf parse-buf)
 	  (set-process-sentinel proc 'notmuch-search-process-sentinel)
 	  (set-process-filter proc 'notmuch-search-process-filter)
 	  (set-process-query-on-exit-flag proc nil))))

test/emacs

@@ -36,7 +36,6 @@ test_emacs '(notmuch-search "tag:inbox")
 test_expect_equal_file OUTPUT $EXPECTED/notmuch-search-tag-inbox
 
 test_begin_subtest "Incremental parsing of search results"
-test_subtest_known_broken
 test_emacs "(ad-enable-advice 'notmuch-search-process-filter 'around 'pessimal)
 	    (ad-activate 'notmuch-search-process-filter)
 	    (notmuch-search \"tag:inbox\")