From: Mark Walters Date: Sat, 25 May 2013 08:59:28 +0000 (+0100) Subject: Re: [PATCH 4/5] emacs: Streaming S-expression parser X-Git-Url: http://git.tremily.us/?a=commitdiff_plain;h=b564f691d91f52dd81dfaa7d1d5ea2ddf545a35e;p=notmuch-archives.git Re: [PATCH 4/5] emacs: Streaming S-expression parser --- diff --git a/94/f7f340780892e36534dd3d0b56331573e98a9d b/94/f7f340780892e36534dd3d0b56331573e98a9d new file mode 100644 index 000000000..604e7ec42 --- /dev/null +++ b/94/f7f340780892e36534dd3d0b56331573e98a9d @@ -0,0 +1,389 @@ +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 47ACD431FAF + for ; Sat, 25 May 2013 01:59:51 -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 P3PKygnSq2pT for ; + Sat, 25 May 2013 01:59:42 -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 E9E0C431FAE + for ; Sat, 25 May 2013 01:59:41 -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 1UgAK2-0000s6-KK; Sat, 25 May 2013 09:59:36 +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 1UgAK1-0003EC-UM; Sat, 25 May 2013 09:59:30 +0100 +From: Mark Walters +To: Austin Clements , notmuch@notmuchmail.org +Subject: Re: [PATCH 4/5] emacs: Streaming S-expression parser +In-Reply-To: <87ip2b3moz.fsf@awakening.csail.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> +User-Agent: Notmuch/0.14+255~gff3cc55 (http://notmuchmail.org) Emacs/23.4.1 + (x86_64-pc-linux-gnu) +Date: Sat, 25 May 2013 09:59:28 +0100 +Message-ID: <8761y7ph8v.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: beeb5d5746d0b2c707bc2474617b0c6f (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 +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: Sat, 25 May 2013 08:59:51 -0000 + + +Hi + +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. + +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? + +>> 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 functions +>>> 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 for +> indicating private symbols. + +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. + +> +>>> +(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."? + +Yes that would be good: or perhaps "like it does in all other code +paths". + +Best wishes=20 + +Mark + +> +>>> + ;; 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-buffer) +>>> + "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: +>>> --=20 +>>> 1.7.10.4