From: Mark Walters Date: Mon, 27 May 2013 20:58:27 +0000 (+0100) Subject: Re: [PATCH 4/5] emacs: Streaming S-expression parser X-Git-Url: http://git.tremily.us/?a=commitdiff_plain;h=d9eb4ba221db80b61ba921d26b2910d97735bb52;p=notmuch-archives.git Re: [PATCH 4/5] emacs: Streaming S-expression parser --- diff --git a/7b/30904a898ebf1c9a55791158c6088c04730bc5 b/7b/30904a898ebf1c9a55791158c6088c04730bc5 new file mode 100644 index 000000000..932e9ef02 --- /dev/null +++ b/7b/30904a898ebf1c9a55791158c6088c04730bc5 @@ -0,0 +1,423 @@ +Return-Path: +X-Original-To: notmuch@notmuchmail.org +Delivered-To: notmuch@notmuchmail.org +Received: from localhost (localhost [127.0.0.1]) + by olra.theworths.org (Postfix) with ESMTP id 95D72431FAF + for ; Mon, 27 May 2013 13:58:40 -0700 (PDT) +X-Virus-Scanned: Debian amavisd-new at olra.theworths.org +X-Spam-Flag: NO +X-Spam-Score: -1.098 +X-Spam-Level: +X-Spam-Status: No, score=-1.098 tagged_above=-999 required=5 + tests=[DKIM_ADSP_CUSTOM_MED=0.001, FREEMAIL_FROM=0.001, + NML_ADSP_CUSTOM_MED=1.2, RCVD_IN_DNSWL_MED=-2.3] autolearn=disabled +Received: from olra.theworths.org ([127.0.0.1]) + by localhost (olra.theworths.org [127.0.0.1]) (amavisd-new, port 10024) + with ESMTP id 92L5TPlETU+D for ; + Mon, 27 May 2013 13:58:34 -0700 (PDT) +Received: from mail2.qmul.ac.uk (mail2.qmul.ac.uk [138.37.6.6]) + (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) + (No client certificate requested) + by olra.theworths.org (Postfix) with ESMTPS id 8E277431FAE + for ; Mon, 27 May 2013 13:58:33 -0700 (PDT) +Received: from smtp.qmul.ac.uk ([138.37.6.40]) + by mail2.qmul.ac.uk with esmtp (Exim 4.71) + (envelope-from ) + id 1Uh4Uv-00042P-GJ; Mon, 27 May 2013 21:58:30 +0100 +Received: from 93-97-24-31.zone5.bethere.co.uk ([93.97.24.31] helo=localhost) + by smtp.qmul.ac.uk with esmtpsa (TLSv1:AES128-SHA:128) (Exim 4.71) + (envelope-from ) + id 1Uh4Uu-0000Wd-Sp; Mon, 27 May 2013 21:58:29 +0100 +From: Mark Walters +To: Austin Clements +Subject: Re: [PATCH 4/5] emacs: Streaming S-expression parser +In-Reply-To: <20130527190436.GS5999@mit.edu> +References: <1368851472-5382-1-git-send-email-amdragon@mit.edu> + <1368851472-5382-5-git-send-email-amdragon@mit.edu> + <87fvxg2wc4.fsf@qmul.ac.uk> + <87ip2b3moz.fsf@awakening.csail.mit.edu> + <8761y7ph8v.fsf@qmul.ac.uk> <20130527190436.GS5999@mit.edu> +User-Agent: Notmuch/0.14+255~gff3cc55 (http://notmuchmail.org) Emacs/23.4.1 + (i486-pc-linux-gnu) +Date: Mon, 27 May 2013 21:58:27 +0100 +Message-ID: <8761y4jg24.fsf@qmul.ac.uk> +MIME-Version: 1.0 +Content-Type: text/plain; charset=utf-8 +Content-Transfer-Encoding: quoted-printable +X-Sender-Host-Address: 93.97.24.31 +X-QM-SPAM-Info: Sender has good ham record. :) +X-QM-Body-MD5: bd9b0ce573f2b98574890d6b736b23f0 (of first 20000 bytes) +X-SpamAssassin-Score: -0.1 +X-SpamAssassin-SpamBar: / +X-SpamAssassin-Report: The QM spam filters have analysed this message to + determine if it is + spam. We require at least 5.0 points to mark a message as spam. + This message scored -0.1 points. + Summary of the scoring: + * 0.0 FREEMAIL_FROM Sender email is commonly abused enduser mail + provider * (markwalters1009[at]gmail.com) + * -0.1 AWL AWL: From: address is in the auto white-list +X-QM-Scan-Virus: ClamAV says the message is clean +Cc: notmuch@notmuchmail.org +X-BeenThere: notmuch@notmuchmail.org +X-Mailman-Version: 2.1.13 +Precedence: list +List-Id: "Use and development of the notmuch mail system." + +List-Unsubscribe: , + +List-Archive: +List-Post: +List-Help: +List-Subscribe: , + +X-List-Received-Date: Mon, 27 May 2013 20:58:40 -0000 + + +Austin Clements writes: + +> Quoth Mark Walters on May 25 at 9:59 am: +>>=20 +>> Hi +>>=20 +>> On Wed, 22 May 2013, Austin Clements wrote: +>> > On Tue, 21 May 2013, Mark Walters wrote: +>> >> Hi +>> >> +>> >> This patch looks good to me. Some minor comments below. +>> > +>> > Some minor replies below. +>> > +>> > In building some other code on top of this, I found an interesting (but +>> > easy to fix) interface bug. Currently, the interface is designed as if +>> > it doesn't matter what buffer these functions are called from, however, +>> > because they move point and expect this point motion to persist, it's +>> > actually not safe to call this interface unless the caller is in the +>> > right buffer anyway. For example, if the buffer is selected in a +>> > window, the with-current-buffer in the parser functions will actually +>> > move a *temporary* point, meaning that the only way the caller can +>> > discover the new point is to first select the buffer for itself. I can +>> > think of two solutions: 1) maintain our own mark for the parser's +>> > current position or 2) tweak the doc strings and code so that it reads +>> > from the current buffer. 1 keeps the interface the way it's currently +>> > documented, but complicates the parser implementation and interface and +>> > doesn't simplify the caller. 2 simplifies the parser and it turns out +>> > all callers already satisfy the requirement. +>>=20 +>> I am confused by this: the docs strings for json/sexp-parse-partial-list +>> both say something like "Parse a partial JSON list from current buffer"? +>> Or do you mean the with-current-buffer in notmuch-search-process-filter? +> +> I was referring to the lower level parser, which effectively has the +> same requirement but isn't documented to and has code that pointlessly +> tries to track the parsing buffer (I consider +> json/sexp-parse-partial-list to be a helper). In fact, one reason the +> lower level parser didn't choke is because right now we only use it +> through json/sexp-parse-partial-list, which requires that it be called +> from the right buffer. + +Right I think I see. I am definitely happy with insisting on being +called from the correct buffer in the low level code. + +>> >> On Sat, 18 May 2013, Austin Clements wrote: +>> >>> This provides the same interface as the streaming JSON parser, but +>> >>> reads S-expressions incrementally. The only difference is that the +>> >>> `notmuch-sexp-parse-partial-list' helper does not handle interleaved +>> >>> error messages (since we now have the ability to separate these out = +at +>> >>> the invocation level), so it no longer takes an error function and +>> >>> does not need to do the horrible resynchronization that the JSON +>> >>> parser had to. +>> >>> +>> >>> Some implementation improvements have been made over the JSON parser. +>> >>> This uses a vector instead of a list for the parser data structure, +>> >>> since this allows faster access to elements (and modern versions of +>> >>> Emacs handle storage of small vectors efficiently). Private functio= +ns +>> >>> follow the "prefix--name" convention. And the implementation is much +>> >>> simpler overall because S-expressions are much easier to parse. +>> >>> --- +>> >>> emacs/Makefile.local | 1 + +>> >>> emacs/notmuch-parser.el | 212 ++++++++++++++++++++++++++++++++++++= ++++++++++++ +>> >>> 2 files changed, 213 insertions(+) +>> >>> create mode 100644 emacs/notmuch-parser.el +>> >>> +>> >>> diff --git a/emacs/Makefile.local b/emacs/Makefile.local +>> >>> index 456700a..a910aff 100644 +>> >>> --- a/emacs/Makefile.local +>> >>> +++ b/emacs/Makefile.local +>> >>> @@ -3,6 +3,7 @@ +>> >>> dir :=3D emacs +>> >>> emacs_sources :=3D \ +>> >>> $(dir)/notmuch-lib.el \ +>> >>> + $(dir)/notmuch-parser.el \ +>> >>> $(dir)/notmuch.el \ +>> >>> $(dir)/notmuch-query.el \ +>> >>> $(dir)/notmuch-show.el \ +>> >>> diff --git a/emacs/notmuch-parser.el b/emacs/notmuch-parser.el +>> >>> new file mode 100644 +>> >>> index 0000000..1b7cf64 +>> >>> --- /dev/null +>> >>> +++ b/emacs/notmuch-parser.el +>> >>> @@ -0,0 +1,212 @@ +>> >>> +;; notmuch-parser.el --- streaming S-expression parser +>> >>> +;; +>> >>> +;; Copyright =C2=A9 Austin Clements +>> >>> +;; +>> >>> +;; This file is part of Notmuch. +>> >>> +;; +>> >>> +;; Notmuch is free software: you can redistribute it and/or modify = +it +>> >>> +;; under the terms of the GNU General Public License as published by +>> >>> +;; the Free Software Foundation, either version 3 of the License, or +>> >>> +;; (at your option) any later version. +>> >>> +;; +>> >>> +;; Notmuch is distributed in the hope that it will be useful, but +>> >>> +;; WITHOUT ANY WARRANTY; without even the implied warranty of +>> >>> +;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +>> >>> +;; General Public License for more details. +>> >>> +;; +>> >>> +;; You should have received a copy of the GNU General Public License +>> >>> +;; along with Notmuch. If not, see . +>> >>> +;; +>> >>> +;; Authors: Austin Clements +>> >>> + +>> >>> +(require 'cl) +>> >>> + +>> >>> +(defun notmuch-sexp-create-parser (buffer) +>> >>> + "Return a streaming S-expression parser that reads from BUFFER. +>> >>> + +>> >>> +This parser is designed to incrementally read an S-expression +>> >>> +whose structure is known to the caller. Like a typical +>> >>> +S-expression parsing interface, it provides a function to read a +>> >>> +complete S-expression from the input. However, it extends this +>> >>> +with an additional function that requires the next value in the +>> >>> +input to be a list and descends into it, allowing its elements to +>> >>> +be read one at a time or further descended into. Both functions +>> >>> +can return 'retry to indicate that not enough input is available. +>> >>> + +>> >>> +The parser always consumes input from BUFFER's point. Hence, the +>> >>> +caller is allowed to delete any data before point and may +>> >>> +resynchronize after an error by moving point." +>> >>> + +>> >>> + (vector 'notmuch-sexp-parser +>> >>> + buffer +>> >>> + ;; List depth +>> >>> + 0 +>> >>> + ;; Partial parse position marker +>> >>> + nil +>> >>> + ;; Partial parse state +>> >>> + nil)) +>> >>> + +>> >>> +(defmacro notmuch-sexp--buffer (sp) `(aref ,sp 1)) +>> >>> +(defmacro notmuch-sexp--depth (sp) `(aref ,sp 2)) +>> >>> +(defmacro notmuch-sexp--partial-pos (sp) `(aref ,sp 3)) +>> >>> +(defmacro notmuch-sexp--partial-state (sp) `(aref ,sp 4)) +>> >> +>> >> Why the double hyphen --? Is it a name-space or some convention? +>> > +>> > More specifically, this seems to be the most common Elisp convention f= +or +>> > indicating private symbols. +>>=20 +>> Ok. If we are keeping the json parser it might be worth making it follow +>> the same convention but as it is purely internal it's probably not worth +>> bothering. +> +> Yeah, it is purely internal. Also, I was planning to remove the JSON +> parser (or maybe move it somewhere else? I feel bad deleting that +> much perfectly functional code, though of course git will keep it in +> perpetuity). + +Yes it does seem a shame to lose it. It would be nice to find a better +home for it then hiding in a git tree. + +Best wishes + +Mark + + +>> >>> +(defun notmuch-sexp-read (sp) +>> >>> + "Consume and return the value at point in SP's buffer. +>> >>> + +>> >>> +Returns 'retry if there is insufficient input to parse a complete +>> >>> +value (though it may still move point over whitespace). If the +>> >>> +parser is currently inside a list and the next token ends the +>> >>> +list, this moves point just past the terminator and returns 'end. +>> >>> +Otherwise, this moves point to just past the end of the value and +>> >>> +returns the value." +>> >>> + +>> >>> + (with-current-buffer (notmuch-sexp--buffer sp) +>> >>> + (skip-chars-forward " \n\r\t") +>> >>> + (cond ((eobp) 'retry) +>> >>> + ((=3D (char-after) ?\)) +>> >>> + ;; We've reached the end of a list +>> >>> + (if (=3D (notmuch-sexp--depth sp) 0) +>> >>> + ;; .. but we weren't in a list. Let read signal the +>> >>> + ;; error. +>> >>> + (read (current-buffer)) +>> >> +>> >> Why is good for read to signal the error rather than us doing it? +>> > +>> > This ensures the syntax error handling and signal behavior of +>> > notmuch-sexp-read is identical in every way to a regular read call. +>> > Maybe the comment should read "Let read signal the error like we do in +>> > all other code paths."? +>>=20 +>> Yes that would be good: or perhaps "like it does in all other code +>> paths". +> +> Sure. +> +>> Best wishes=20 +>>=20 +>> Mark +>>=20 +>> > +>> >>> + ;; Go up a level and return an end token +>> >>> + (decf (notmuch-sexp--depth sp)) +>> >>> + (forward-char) +>> >>> + 'end)) +>> >>> + ((=3D (char-after) ?\() +>> >>> + ;; We're at the beginning of a list. If we haven't started +>> >>> + ;; a partial parse yet, attempt to read the list in its +>> >>> + ;; entirety. If this fails, or we've started a partial +>> >>> + ;; parse, extend the partial parse to figure out when we +>> >>> + ;; have a complete list. +>> >>> + (catch 'return +>> >>> + (when (null (notmuch-sexp--partial-state sp)) +>> >>> + (let ((start (point))) +>> >>> + (condition-case nil +>> >>> + (throw 'return (read (current-buffer))) +>> >>> + (end-of-file (goto-char start))))) +>> >>> + ;; Extend the partial parse +>> >>> + (let (is-complete) +>> >>> + (save-excursion +>> >>> + (let* ((new-state (parse-partial-sexp +>> >>> + (or (notmuch-sexp--partial-pos sp) (point)) +>> >>> + (point-max) 0 nil +>> >>> + (notmuch-sexp--partial-state sp))) +>> >>> + ;; A complete value is available if we've +>> >>> + ;; reached depth 0. +>> >>> + (depth (first new-state))) +>> >>> + (assert (>=3D depth 0)) +>> >>> + (if (=3D depth 0) +>> >>> + ;; Reset partial parse state +>> >>> + (setf (notmuch-sexp--partial-state sp) nil +>> >>> + (notmuch-sexp--partial-pos sp) nil +>> >>> + is-complete t) +>> >>> + ;; Update partial parse state +>> >>> + (setf (notmuch-sexp--partial-state sp) new-state +>> >>> + (notmuch-sexp--partial-pos sp) (point-marker))))) +>> >>> + (if is-complete +>> >>> + (read (current-buffer)) +>> >>> + 'retry)))) +>> >>> + (t +>> >>> + ;; Attempt to read a non-compound value +>> >>> + (let ((start (point))) +>> >>> + (condition-case nil +>> >>> + (let ((val (read (current-buffer)))) +>> >>> + ;; We got what looks like a complete read, but if +>> >>> + ;; we reached the end of the buffer in the process, +>> >>> + ;; we may not actually have all of the input we +>> >>> + ;; need (unless it's a string, which is delimited). +>> >>> + (if (or (stringp val) (not (eobp))) +>> >>> + val +>> >>> + ;; We can't be sure the input was complete +>> >>> + (goto-char start) +>> >>> + 'retry)) +>> >>> + (end-of-file +>> >>> + (goto-char start) +>> >>> + 'retry))))))) +>> >>> + +>> >>> +(defun notmuch-sexp-begin-list (sp) +>> >>> + "Parse the beginning of a list value and enter the list. +>> >>> + +>> >>> +Returns 'retry if there is insufficient input to parse the +>> >>> +beginning of the list. If this is able to parse the beginning of +>> >>> +a list, it moves point past the token that opens the list and +>> >>> +returns t. Later calls to `notmuch-sexp-read' will return the +>> >>> +elements inside the list. If the input in buffer is not the +>> >>> +beginning of a list, throw invalid-read-syntax." +>> >>> + +>> >>> + (with-current-buffer (notmuch-sexp--buffer sp) +>> >>> + (skip-chars-forward " \n\r\t") +>> >>> + (cond ((eobp) 'retry) +>> >>> + ((=3D (char-after) ?\() +>> >>> + (forward-char) +>> >>> + (incf (notmuch-sexp--depth sp)) +>> >>> + t) +>> >>> + (t +>> >>> + ;; Skip over the bad character like `read' does +>> >>> + (forward-char) +>> >>> + (signal 'invalid-read-syntax (list (string (char-before)))))))) +>> >>> + +>> >>> +(defun notmuch-sexp-eof (sp) +>> >>> + "Signal an error if there is more data in SP's buffer. +>> >>> + +>> >>> +Moves point to the beginning of any trailing data or to the end +>> >>> +of the buffer if there is only trailing whitespace." +>> >>> + +>> >>> + (with-current-buffer (notmuch-sexp--buffer sp) +>> >>> + (skip-chars-forward " \n\r\t") +>> >>> + (unless (eobp) +>> >>> + (error "Trailing garbage following expression")))) +>> >>> + +>> >>> +(defvar notmuch-sexp--parser nil +>> >>> + "The buffer-local notmuch-sexp-parser instance. +>> >>> + +>> >>> +Used by `notmuch-sexp-parse-partial-list'.") +>> >>> + +>> >>> +(defvar notmuch-sexp--state nil +>> >>> + "The buffer-local `notmuch-sexp-parse-partial-list' state.") +>> >>> + +>> >>> +(defun notmuch-sexp-parse-partial-list (result-function result-buff= +er) +>> >>> + "Incrementally parse an S-expression list from the current buffer. +>> >>> + +>> >>> +This function consume an S-expression list from the current +>> >> +>> >> consumes +>> > +>> > Oops, yes. +>> > +>> >>> +buffer, applying RESULT-FUNCTION in RESULT-BUFFER to each +>> >>> +complete value in the list. It operates incrementally and should +>> >>> +be called whenever the input buffer has been extended with +>> >>> +additional data. The caller just needs to ensure it does not +>> >>> +move point in the input buffer." +>> >>> + +>> >>> + ;; Set up the initial state +>> >>> + (unless (local-variable-p 'notmuch-sexp--parser) +>> >>> + (set (make-local-variable 'notmuch-sexp--parser) +>> >>> + (notmuch-sexp-create-parser (current-buffer))) +>> >>> + (set (make-local-variable 'notmuch-sexp--state) 'begin)) +>> >>> + (let (done) +>> >>> + (while (not done) +>> >>> + (case notmuch-sexp--state +>> >>> + (begin +>> >>> + ;; Enter the list +>> >>> + (if (eq (notmuch-sexp-begin-list notmuch-sexp--parser) 'retry) +>> >>> + (setq done t) +>> >>> + (setq notmuch-sexp--state 'result))) +>> >>> + (result +>> >>> + ;; Parse a result +>> >>> + (let ((result (notmuch-sexp-read notmuch-sexp--parser))) +>> >>> + (case result +>> >>> + (retry (setq done t)) +>> >>> + (end (setq notmuch-sexp--state 'end)) +>> >>> + (t (with-current-buffer result-buffer +>> >>> + (funcall result-function result)))))) +>> >>> + (end +>> >>> + ;; Any trailing data is unexpected +>> >>> + (notmuch-sexp-eof notmuch-sexp--parser) +>> >>> + (setq done t))))) +>> >>> + ;; Clear out what we've parsed +>> >>> + (delete-region (point-min) (point))) +>> >>> + +>> >>> +(provide 'notmuch-parser) +>> >>> + +>> >>> +;; Local Variables: +>> >>> +;; byte-compile-warnings: (not cl-functions) +>> >>> +;; End: