Next: Introduction [Contents]
Next: m-buffer, Previous: Manipulate the Contents of Emacs Buffers, Up: Manipulate the Contents of Emacs Buffers [Contents]
m-buffer provides functions for accessing and manipulating the contents of an Emacs buffer. While Emacs already provides these features, m-buffer provides a higher-level interaction. It achieves this in several ways: many of the functions are list-orientated, so avoiding the need for iteration; it avoids the use of global emacs state whenever it can be avoided, so avoiding side-effects; and it provides a large library of functions supporting common operations.
Core usage of buffer m-buffer is simple. For example, the following code returns a list of all matches to the regexp "m-buffer" in the ‘current-buffer‘.
(m-buffer-match (current-buffer) "m-buffer")
m-buffer is also expanding. Other parts of m-buffer provide stateless interaction with the existing buffer; for example, we can use the following to fetch the point of any buffer:
(m-buffer-at-point buffer)
These functions can help greatly when writing code which operates on two or more buffers. It is also possible to check whether the status of a location – either a buffer and position or a marker. For example, these calls are equivalent to ‘eolp‘.
(m-buffer-at-eolp buffer position) (m-buffer-at-eolp marker)
Up: Introduction [Contents]
‘m-buffer’ is a work in progress, but much of it is now stable and the interface should change only in forward-compatible ways for 1.0 release. The individual files have statements about their stability.
Next: m-buffer-at, Previous: Introduction, Up: Manipulate the Contents of Emacs Buffers [Contents]
m-buffer.el provides list-orientated search both for any regexp and standard regexps, as well as the ability to do things with these matches: replace, add overlays or text-properties or, most generically of all, call any function on matches.
Next: Commentary, Up: m-buffer [Contents]
This file is not part of Emacs
Author: Phillip Lord <phillip.lord@russet.org.uk> Maintainer: Phillip Lord <phillip.lord@russet.rg.uk> Version: 0.16 Package-Requires: ((seq "2.14"))
The contents of this file are subject to the GPL License, Version 3.0.
Copyright (C) 2014-2024 Free Software Foundation, Inc.
This program 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
This program 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 this program. If not, see http://www.gnu.org/licenses/.
This file provides a set of list-oriented functions for operating over the contents of buffers, mostly revolving around regexp searching, and regions. They avoid the use of looping, manipulating global state with ‘match-data’. Many high-level functions exist for matching sentences, lines and so on.
Functions are generally purish: i.e. that is those functions which do change state, by for example replacing text or adding overlays, should only change state in one way; they will not affect point, current buffer, match data or so forth.
Likewise to protect against changes in state, markers are used rather than integer positions. This means that it is possible, for example, to search for regexp matches and then replace them all without the earlier replacements invalidating the location of the later ones. Otherwise replacements need to be made in reverse order. This can have implications for performance, so m-buffer also provides functions for making markers nil; there are also macros which help manage markers in ‘m-buffer-macro’.
Where possible, functions share interfaces. So most of the match functions take a list of "match" arguments, either position or as a plist, which avoids using lots of ‘nil’ arguments. Functions operating on matches take a list of ‘match-data’ as returned by the match functions, making it easy to chain matches.
This file is documented using lentic.el. Use lentic-server to view.
Next: Code, Previous: Commentary, Up: m-buffer [Contents]
m-buffer.el is now stable and is expected to change only in forward-compatible ways.
(require 'seq) (require 'm-buffer-macro)
Next: Match Data Manipulation Functions, Up: Code [Contents]
We first provide a single match function, ‘m-bufffer-match’ which converts between Emacs’ stateful matching and a more sequence-oriented interface.
This function also defines the "match" arguments which are a standard set of arguments used throughout this package.
(defun m-buffer-match (&rest match)
"Return a list of all `match-data' for MATCH.
MATCH may be of the forms:
BUFFER REGEXP &optional MATCH-OPTIONS
WINDOW REGEXP &optional MATCH-OPTIONS
MATCH-OPTIONS
If BUFFER is given, search this buffer. If WINDOW is given search
the visible window. MATCH-OPTIONS is a plist with any of the
following keys:
:buffer -- the buffer to search
:regexp -- the regexp to search with
:begin -- the start of the region to search -- default point min
:end -- the end of the region to search -- default point max
:post-match -- function called after a match -- default nil
:widen -- if true, widen buffer first -- default nil
:case-fold-search value of `case-fold-search' during search.
If :default accept the current buffer-local value
:numeric -- if true, return integers not markers
If options are expressed in two places, the plist form takes
precedence over positional args. So calling with both a first
position buffer and a :buffer arg will use the second. Likewise,
if a window is given as first arg and :end is given, then
the :end value will be used.
REGEXP should advance point (i.e. not be zero-width) or the
function will loop infinitely. POST-MATCH can be used to avoid
this. The buffer is searched forward."
(apply #'m-buffer--match-1
(m-buffer--normalize-args match)))
The match function is actually implemented here in the ‘m-buffer–match-1’ function, with positional arguments.
(defun m-buffer--match-1 (buffer regexp begin end
post-match widen cfs
numeric)
"Return a list of `match-data'.
This is an internal function: please prefer `m-buffer-match'.
BUFFER -- the buffer.
REGEXP -- the regexp.
BEGIN -- the start of the region to search
END -- the end of the region to search
POST-MATCH -- function to run after each match
POST-MATCH is useful for zero-width matches which will otherwise
cause infinite loop. The buffer is searched forward. POST-MATCH
return can also be used to terminate the matching by returning nil.
WIDEN -- call widen first.
CFS -- Non-nil if searches and matches should ignore case.
NUMERIC -- Non-nil if we should return integers not markers."
We start by saving everything to ensure that we do not pollute the global state. This means match-data, point, narrowing and current buffer! Hopefully this is all the global state that exists and that we are changing.
(with-current-buffer
buffer
(save-match-data
(save-excursion
(save-restriction
(when widen (widen))
This let form is doing a number of things. It sets up a dynamic binding for ‘case-fold-search’ (which works even though we are using lexical binding), ensures a non-nil value for ‘end-bound’ and defines a sentinal value that ‘post-match-return’ can use to end early.
(let ((rtn nil)
(post-match-return t)
(end-bound (or end (point-max)))
;; over-ride default if necessary
(case-fold-search
(if (eq :default cfs)
case-fold-search
cfs)))
We start at the beginning. There was no particularly good reason for this, and it would have made just as much sense to go backward.
(goto-char
(or begin
(point-min)))
(while
(and
The original purpose for ‘post-match-return’ was for zero-width matches – these do not advance point beyond their end, so the while loop never terminates. Unfortunately, avoiding this depends on the regexp being called, so we provide the most general solution of all.
As well as this, we check the return value of ‘post-match-return’, so as well as advancing ‘point’ by side-effect, we can also use it to terminate the look at any point that we want; for example, we can terminate after the first match which feels more efficient than searching the whole buffer then taking the first match.
post-match-return ;; we need to check we are less than the end-bound ;; or re-search-forward will break (<= (point) end-bound) (re-search-forward regexp end-bound t))
Store the ‘match-data’ in a backward list, run post-match. Finally, reverse and terminate.
(setq rtn
(cons
(if numeric
(m-buffer-marker-to-pos-nil
(match-data))
(match-data))
rtn))
(when post-match
(setq post-match-return (funcall post-match))))
(reverse rtn)))))))
This method implements the argument list processing. I find this interface fairly attractive to use since it takes the two "main" arguments – buffer and regexp – as positional args optionally, and everything else as keywords. The use of keywords is pretty much essential as have eight arguments most of which are not essential.
This is fairly close to the logic provided by ‘cl-defun’ which I wasn’t aware of when I wrote this. However ‘cl-defun’ does not allow optional arguments before keyword arguments – all the optional arguments have to be given if we are to use keywords.
(defun m-buffer--normalize-args (match-with)
"Manipulate args into a standard form and return as a list.
MATCH-WITH are these args. This is an internal function."
(let* (
;; split up into keyword and non keyword limits
(args
(seq-take-while
(lambda (x) (not (keywordp x)))
match-with))
(pargs
(seq-drop-while
(lambda (x) (not (keywordp x)))
match-with))
;; sort actual actual parameters
(first (car args))
;; buffer may be first
(buffer
(or (plist-get pargs :buffer)
(and (bufferp first) first)))
;; or window may be first
(window
(or (plist-get pargs :window)
(and (windowp first) first)))
;; regexp always comes second
(regexp
(or (plist-get pargs :regexp)
(nth 1 args)))
;; begin depends on other arguments
(begin
(or (plist-get pargs :begin)
(and window (window-start window))))
;; end depends on other arguments
(end
(or (plist-get pargs :end)
(and window (window-end window))))
;; pm
(post-match
(plist-get pargs :post-match))
;; widen
(widen
(plist-get pargs :widen))
;; case-fold-search this needs to overwrite the buffer contents iff
;; set, otherwise be ignored, so we need to distinguish a missing
;; property and a nil one
(cfs
(if (plist-member pargs :case-fold-search)
(plist-get pargs :case-fold-search)
:default))
;; numeric
(numeric
(plist-get pargs :numeric)))
(list buffer regexp begin end post-match widen cfs numeric)))
Finally, this function provides a link between the match function, and the match manipulation functions. We can either choose to match once against a set of arguments and then apply multiple manipulations on the returned match data. Or just use the match manipulation function directly.
The first version of ‘m-buffer’ did not include this but it required lots of nested calls which seem inconvenient.
(m-buffer-match-manipulate (m-buffer-match (current-buffer) "hello"))
I think that convienience is worth the overhead.
(defun m-buffer-ensure-match (&rest match)
"Ensure that we have MATCH data.
If a single arg, assume it is match data and return. If multiple
args, assume they are of the form accepted by
`m-buffer-match'."
(cond
;; we have match data
((= 1 (length match))
(car match))
((< 1 (length match))
(apply #'m-buffer-match match))
(t
(error "Invalid arguments"))))
Next: Match Utility and Predicates, Previous: Regexp Matching, Up: Code [Contents]
These functions manipulate lists of either match-data or match arguments in some way.
(defun m-buffer-buffer-for-match (match-data)
"Given some MATCH-DATA return the buffer for that data."
(marker-buffer (caar match-data)))
(defun m-buffer-match-nth-group (n match-data)
"Fetch the Nth group from MATCH-DATA."
(seq-map
(lambda (m)
(let ((drp
(seq-drop m (* 2 n))))
(list
(car drp) (cadr drp))))
match-data))
(defun m-buffer-match-begin-n (n &rest match)
"Return markers to the start of the Nth group in MATCH.
MATCH may be of any form accepted by `m-buffer-ensure-match'. Use
`m-buffer-nil-marker' after the markers have been finished with
or they will slow future use of the buffer until garbage collected."
(seq-map
(lambda (m)
(nth
(* 2 n) m))
(apply #'m-buffer-ensure-match match)))
(defun m-buffer-match-begin-n-pos (n &rest match)
"Return positions of the start of the Nth group in MATCH.
MATCH may be of any form accepted by `m-buffer-ensure-match'. If
`match-data' is passed markers will be set to nil after this
function. See `m-buffer-nil-marker' for details."
(m-buffer-marker-to-pos-nil
(apply #'m-buffer-match-begin-n
n match)))
(defun m-buffer-match-begin (&rest match)
"Return a list of markers to the start of MATCH.
MATCH may of any form accepted by `m-buffer-ensure-match'. Use
`m-buffer-nil-marker' after the markers have been used or they
will slow future changes to the buffer."
(apply #'m-buffer-match-begin-n 0 match))
(defun m-buffer-match-begin-pos (&rest match)
"Return a list of positions at the start of matcher.
MATCH may be of any form accepted by `m-buffer-ensure-match'.
If `match-data' is passed markers will be set to nil after this
function. See `m-buffer-nil-marker' for details."
(apply #'m-buffer-match-begin-n-pos 0 match))
(defun m-buffer-match-end-n (n &rest match)
"Return markers to the end of the match to the Nth group.
MATCH may be of any form accepted by `m-buffer-ensure-match'.
If `match-data' is passed markers will be set to nil after this
function. See `m-buffer-nil-marker' for details."
(seq-map
(lambda (m)
(nth
(+ 1 (* 2 n))
m))
(apply #'m-buffer-ensure-match match)))
(defun m-buffer-match-end-n-pos (n &rest match)
"Return positions of the end Nth group of MATCH.
MATCH may be of any form accepted by `m-buffer-ensure-match'.
If `match-data' is passed markers will be set to nil after this
function. See `m-buffer-nil-marker' for details."
(m-buffer-marker-to-pos-nil
(apply #'m-buffer-match-end-n-pos
n match)))
(defun m-buffer-match-end (&rest match)
"Return a list of markers to the end of MATCH to regexp in buffer.
MATCH may be of any form accepted by `m-buffer-ensure-match'. Use
`m-buffer-nil-marker' after the markers have been used or they
will slow future changes to the buffer."
(apply #'m-buffer-match-end-n 0 match))
(defun m-buffer-match-end-pos (&rest match)
"Return a list of positions to the end of the match.
MATCH may be of any form accepted by `m-buffer-ensure-match'.
If `match-data' is passed markers will be set to nil after this
function. See `m-buffer-nil-marker' for details."
(m-buffer-marker-to-pos-nil
(apply #'m-buffer-match-end match)))
Next: Marker manipulation functions, Previous: Match Data Manipulation Functions, Up: Code [Contents]
Some predicates and the ability to subtract to lists of matches from each other. This makes up for limitations in Emacs regexp which can’t do "match x but not y".
(defun m-buffer-match-equal (m n)
"Return true if M and N are cover the same region.
Matches are equal if they match the same region; subgroups are
ignored."
(and
(equal
(car m)
(car n))
(equal
(cadr m)
(cadr n))))
A nice simple implementation for the general purpose solution. Unfortunately, performance sucks, running in quadratic time.
(defun m-buffer-match-subtract (m n)
"Remove from M any match in N.
Matches are equivalent if overall they match the same
area; subgroups are ignored.
See also `m-buffer-match-exact-subtract' which often
runs faster but has some restrictions."
(seq-remove
(lambda (o)
(seq-some
(lambda (p)
(m-buffer-match-equal o p))
n))
m))
The ugly and complicated and less general solution. But it runs in linear time.
(defun m-buffer-match-exact-subtract (m n)
"Remove from M any match in N.
Both M and N must be fully ordered, and any element in N must be
in M."
(if n
;; n-eaten contains the remaining elements of n that we haven't tested
;; for yet. We throw them away as we go
(let ((n-eaten n))
(seq-remove
(lambda (o)
(cond
;; n-eaten has been eaten. Check here or later "<" comparison crashes.
((not n-eaten)
;; return nil because we always want things in m now.
nil
)
;; we have a match so throw away the first element of n-eaten
;; which we won't need again.
((m-buffer-match-equal
(car n-eaten) o)
(progn
(setq n-eaten (seq-drop n-eaten 1))
t))
;; we should discard also if n-eaten 1 is less than o because, both
;; are sorted, so we will never match
((<
;; first half of the first match in n-eaten
(caar n-eaten)
;; first half of match
(car o))
(progn
(setq n-eaten (seq-drop n-eaten 1))
t))))
m))
m))
(defun m-buffer-in-match-p (matches position)
"Returns true is any of MATCHES contain POSITION."
(seq-some
(lambda (match)
(and
(<= (car match) position)
(<= position (cadr match))))
matches))
Partition one set of markers by another. This is useful for finding matched pairs of markers.
(defun m-buffer--partition-by-marker(list partition)
"Given LIST, split at markers in PARTITION.
This is the main implementation for `m-buffer-partition-by-marker',
but assumes that partition starts with a very low value (or nil)."
(let* ((p-top (car-safe partition))
(p-val (car-safe (cdr-safe partition)))
(p-fn (lambda (n)
(or (not p-val)
(< n p-val)))))
(when list
(cons
(cons
p-top
(seq-take-while p-fn list))
(m-buffer--partition-by-marker
(seq-drop-while p-fn list)
(cdr partition))))))
(defun m-buffer-partition-by-marker (list partition)
"Given LIST of markers, split at markers in PARTITION.
Returns a list of lists. The first element of each list is nil or
the marker from PARTITION. The rest of the elements are those
elements in LIST which are at the same position or later in the
buffer than the element from PARTITION, but before the next
element from PARTITION.
Both LIST and PARTITION must be sorted."
;; TODO!
(m-buffer--partition-by-marker list (cons nil partition)))
Next: Replace, Delete, Extract, Previous: Match Utility and Predicates, Up: Code [Contents]
These functions do things to markers rather than the areas of the buffers indicated by the markers. This includes transforming between markers and integer positions, and niling markers explicitly, which prevents slow down before garbage collection.
(defun m-buffer-nil-marker (markers)
"Takes a (nested) list of MARKERS and nils them all.
Markers slow buffer movement while they are pointing at a
specific location, until they have been garbage collected. Niling
them prevents this. See Info node `(elisp) Overview of Markers'."
(seq-map
(lambda (marker)
(if (seqp marker)
(m-buffer-nil-marker marker)
(set-marker marker nil)))
markers))
(defun m-buffer-marker-to-pos (markers &optional postnil)
"Transforms a list of MARKERS to a list of positions.
If the markers are no longer needed, set POSTNIL to true, or call
`m-buffer-nil-marker' manually after use to speed future buffer
movement. Or use `m-buffer-marker-to-pos-nil'."
(seq-map
(lambda (marker)
(prog1
(marker-position marker)
(when postnil
(set-marker marker nil))))
markers))
(defun m-buffer-marker-to-pos-nil (markers)
"Transforms a list of MARKERS to a list of positions then nils.
See also `m-buffer-nil-markers'"
(m-buffer-marker-to-pos markers t))
(defun m-buffer-marker-tree-to-pos (marker-tree &optional postnil)
"Transforms a tree of markers to equivalent positions.
MARKER-TREE is the tree.
POSTNIL sets markers to nil afterwards."
(seq-map
(lambda (marker)
(if (seqp marker)
(m-buffer-marker-tree-to-pos marker postnil)
(prog1
(marker-position marker)
(when postnil
(set-marker marker nil)))))
marker-tree))
(defun m-buffer-marker-tree-to-pos-nil (marker-tree)
"Transforms a tree of markers to equivalent positions.
MARKER-TREE is the tree. Markers are niled afterwards."
(m-buffer-marker-tree-to-pos marker-tree t))
(defun m-buffer-marker-clone (marker-tree &optional type)
"Return a clone of MARKER-TREE.
The optional argument TYPE specifies the insertion type. See
`copy-marker' for details."
(seq-map
(lambda (marker)
(if (seqp marker)
(m-buffer-marker-clone marker type)
(copy-marker marker type)))
marker-tree))
(defun m-buffer-pos-to-marker (buffer positions)
"In BUFFER translates a list of POSITIONS to markers."
(seq-map
(lambda (pos)
(set-marker
(make-marker) pos buffer))
positions))
Next: Match Things, Previous: Marker manipulation functions, Up: Code [Contents]
(defun m-buffer-replace-match (match-data replacement
&optional fixedcase literal subexp)
"Given a list of MATCH-DATA, replace with REPLACEMENT.
If FIXEDCASE do not alter the case of the replacement text.
If LITERAL insert the replacement literally.
SUBEXP should be a number indicating the regexp group to replace.
Returns markers to the start and end of the replacement. These
markers are part of MATCH-DATA, so niling them will percolate backward.
See also `replace-match'."
(save-excursion
(seq-map
(lambda (match)
(with-current-buffer
(marker-buffer (car match))
(save-match-data
(set-match-data match)
(replace-match
replacement fixedcase literal nil
(or subexp 0)))))
match-data))
;; we have match-data
(m-buffer-match-nth-group (or subexp 0) match-data))
(defun m-buffer-delete-match (match-data &optional subexp)
"Delete all MATCH-DATA.
SUBEXP should be a number indicating the regexp group to delete.
Returns markers to the start and end of the replacement. These
markers are part of MATCH_DATA, so niling them will percolate backward."
(m-buffer-replace-match match-data "" subexp))
(defun m-buffer-match-string (match-data &optional subexp)
"Return strings for MATCH-DATA optionally of group SUBEXP."
(seq-map
(lambda (match)
(with-current-buffer
(marker-buffer (car match))
(save-match-data
(set-match-data match)
(match-string
(or subexp 0)))))
match-data))
(defun m-buffer-match-string-no-properties (match-data &optional subexp)
"Return strings for MATCH-DATA optionally of group SUBEXP.
Remove all properties from return."
(seq-map
#'substring-no-properties
(m-buffer-match-string
match-data subexp)))
Next: Apply Function to Match, Previous: Replace, Delete, Extract, Up: Code [Contents]
Emacs comes with a set of in-built regexps most of which we use here.
We define ‘m-buffer-apply-join’ first. The reason for this function is that we want to take a list of match arguments and add to with, for instance, a regular expression. We need to add these at the end because most of our functions contain some positional arguments.
(defun m-buffer-apply-join (fn match &rest more-match)
(let*
((args
(seq-take-while
(lambda (x) (not (keywordp x)))
match))
(pargs
(seq-drop-while
(lambda (x) (not (keywordp x)))
match))
(more-keywords
(seq-map
#'car
(seq-partition more-match 2))))
(when
(seq-find
(lambda (keyword)
(plist-member pargs keyword))
more-keywords)
(error
"Match arg contradicts a defined argument."))
(apply fn (append args more-match pargs))))
For the following code, we use Emacs core regexps where possible.
(defun m-buffer-match-page (&rest match)
"Return a list of match data to all pages in MATCH.
MATCH is of form BUFFER-OR-WINDOW MATCH-OPTIONS. See
`m-buffer-match' for further details."
(m-buffer-apply-join 'm-buffer-match
match :regexp page-delimiter))
The ‘paragraph-separate’ regexp can match an empty region, so we need to start each search at the beginning of the next line.
(defun m-buffer-match-paragraph-separate (&rest match) "Return a list of match data to `paragraph-separate' in MATCH. MATCH is of form BUFFER-OR-WINDOW MATCH-OPTIONS. See `m-buffer-match' for futher details." (m-buffer-apply-join 'm-buffer-match match :regexp paragraph-separate :post-match 'm-buffer-post-match-forward-line)) (defvar m-buffer--line-regexp "^.*$" "Regexp to match a line.") (defun m-buffer-match-line (&rest match) "Return a list of match data to all lines. MATCH is of the form BUFFER-OR-WINDOW MATCH-OPTIONS. See `m-buffer-match for further details." (m-buffer-apply-join 'm-buffer-match match :regexp m-buffer--line-regexp :post-match 'm-buffer-post-match-forward-char)) (defun m-buffer-match-line-start (&rest match) "Return a list of match data to all line start. MATCH is of form BUFFER-OR-WINDOW MATCH-OPTIONS. See `m-buffer-match' for further details." (m-buffer-apply-join 'm-buffer-match-begin match :regexp "^" :post-match 'm-buffer-post-match-forward-char)) (defun m-buffer-match-line-end (&rest match) "Return a list of match to line end. MATCH is of form BUFFER-OR-WINDOW MATCH-OPTIONS. See `m-buffer-match' for further details." (m-buffer-apply-join 'm-buffer-match-begin match :regexp "$" :post-match 'm-buffer-post-match-forward-char))
This is the first use of the ‘post-match’ to terminate the loop, and was actually the motivation for adding it. We automatically terminate after the first match by simply returning nil.
(defun m-buffer-match-first (&rest match)
"Return the first match to MATCH.
This matches more efficiently than matching all matches and
taking the car. See `m-buffer-match' for further details of
MATCH."
(m-buffer-apply-join
#'m-buffer-match match
:post-match (lambda () nil)))
(defun m-buffer-match-first-line (&rest match)
"Return a match to the first line of MATCH.
This matches more efficiently than matching all lines and taking
the car. See `m-buffer-match' for further details of MATCH."
(m-buffer-apply-join
'm-buffer-match-first match
:regexp m-buffer--line-regexp))
(defun m-buffer-match-multi (regexps &rest match)
"Incrementally find matches to REGEXPS in MATCH.
Finds the first match to the first element of regexps, then
starting from the end of this match, the first match to the
second element of regexps and so forth. See `m-buffer-match' for
futher details of MATCH."
(when regexps
(let ((first-match
(m-buffer-apply-join
#'m-buffer-match-first
match
:regexp (car regexps))))
(append
first-match
(apply
#'m-buffer-match-multi
(cdr regexps)
(plist-put
match
:begin (car (m-buffer-match-end first-match))))))))
Emacs has a rather inconsistent interface here – suddenly, we have a function rather than a variable for accessing a regexp.
(defun m-buffer-match-sentence-end (&rest match) "Return a list of match to sentence end. MATCH is of the form BUFFER-OR-WINDOW MATCH-OPTIONS. See `m-buffer-match' for further details." (m-buffer-apply-join 'm-buffer-match-begin match :regexp (sentence-end))) (defun m-buffer-match-word (&rest match) "Return a list of match to all words. MATCH is of the form BUFFER-OR-WINDOW MATCH-OPTIONS. See `m-buffer-match' for further details." (m-buffer-apply-join 'm-buffer-match match :regexp "\\\w+")) (defun m-buffer-match-empty-line (&rest match) "Return a list of match to all empty lines. MATCH is of the form BUFFER-OR-WINDOW MATCH-OPTIONS. See `m-buffer-match' for further details." (m-buffer-apply-join 'm-buffer-match match :regexp "^$" :post-match 'm-buffer-post-match-forward-line)) (defun m-buffer-match-non-empty-line (&rest match) "Return a list of match to all non-empty lines. MATCH is fo the form BUFFER-OR-WINDOW MATCH-OPTIONS. See `m-buffer-match' for further details." (m-buffer-apply-join 'm-buffer-match match :regexp "^.+$")) (defun m-buffer-match-whitespace-line (&rest match) "Return match data to all lines with only whitespace characters. Note empty lines are not included. MATCH is of form BUFFER-OR-WINDOW MATCH-OPTIONS. See `m-buffer-match' for further details." (m-buffer-apply-join 'm-buffer-match match :regexp "^\\s-+$"))
I don’t think that there is a way to do this with regexps entirely, so we use substraction.
(defun m-buffer-match-non-whitespace-line (&rest match)
"Return match data to all lines with at least one non-whitespace character.
Note empty lines do not contain any non-whitespace lines.
MATCH is of form BUFFER-OR-WINDOW MATCH-OPTIONS. See
`m-buffer-match' for further details."
(seq-difference
(apply #'m-buffer-match-line match)
(apply #'m-buffer-match-whitespace-line match)))
;; Useful post-match functions
(defun m-buffer-post-match-forward-line ()
"Attempt to move forward one line, return true if success."
(= 0 (forward-line)))
(defun m-buffer-post-match-forward-char ()
"Attempts to move forward one char.
Returns true if succeeds."
(condition-case _e
(progn
(forward-char)
t)
(error 'end-of-buffer
nil)))
Next: Overlay and Property Functions, Previous: Match Things, Up: Code [Contents]
These functions apply another function to some match-data. This is pretty useful generically, but also I use it for many of the following functions.
(defun m-buffer-on-region (fn match-data)
"Apply FN to MATCH-DATA.
FN should take two args, the start and stop of each region.
MATCH-DATA can be any list of lists with two elements (or more)."
(m-buffer-on-region-nth-group fn 0 match-data))
(defun m-buffer-on-region-nth-group (fn n match-data)
"Apply FN to the Nth group of MATCH-DATA.
FN should take two args, the start and stop of each region.
MATCH-DATA can be any list of lists with two elements (or more)."
(seq-map
(lambda (x)
(apply fn x))
(m-buffer-match-nth-group n match-data)))
Previous: Apply Function to Match, Up: Code [Contents]
Adding properties or overlays to match-data. The functionality here somewhat overlaps with ov.el, which I didn’t know about when I wrote this. It generally works over overlays, or regexps, while m-buffer works over match-data.
(defun m-buffer-overlay-match (match-data &optional front-advance rear-advance)
"Return an overlay for all match to MATCH-DATA.
FRONT-ADVANCE and REAR-ADVANCE controls the borders of the
overlay as defined in `make-overlay'. Overlays do not scale that
well, so use `m-buffer-propertize-match' if you intend to make
and keep many of these.
See Info node `(elisp) Overlays' for further information."
(let ((buffer (m-buffer-buffer-for-match match-data)))
(m-buffer-on-region
(lambda (beginning end)
(make-overlay
beginning end buffer
front-advance rear-advance))
match-data)))
(defun m-buffer-add-text-property-match
(match-data properties)
"To MATCH-DATA add PROPERTIES.
See `add-text-property' for details of the format of properties.
Text properties are associated with the text and move with it. See
Info node `(elisp) Text Properties' for further details."
(m-buffer-on-region
(lambda (beginning end)
(add-text-properties beginning end properties))
match-data))
(defun m-buffer-put-text-property-match (match-data property value)
"To MATCH-DATA add PROPERTY wth VALUE.
See `put-text-property' for details of the format of properties.
Text properties are associated with the text and move with it. See
Info node `(elisp) Text Properties' for further details."
(m-buffer-on-region
(lambda (beginning end)
(put-text-property beginning end property value))
match-data))
(defun m-buffer-overlay-face-match (match-data face)
"To MATCH-DATA add FACE to the face property.
This is for use in buffers which do not have function `font-lock-mode'
enabled; otherwise use `m-buffer-overlay-font-lock-face-match'."
(seq-map
(lambda (ovly)
(overlay-put ovly 'face face))
(m-buffer-overlay-match match-data)))
(defun m-buffer-overlay-font-lock-face-match (match-data face)
"To MATCH-DATA add FACE to the face property.
This is for use in buffers which have variable `font-lock-mode' enabled;
otherwise use `m-buffer-overlay-face-match'."
(seq-map
(lambda (ovly)
(overlay-put ovly 'face face))
(m-buffer-overlay-match match-data)))
(defun m-buffer-text-property-face (match-data face)
"To MATCH-DATA apply FACE.
This is for use in buffers which do
not have variable `font-lock-mode' enabled; otherwise use
`m-buffer-text-property-font-lock-face'."
(m-buffer-put-text-property-match match-data
'face face))
(defun m-buffer-text-property-font-lock-face (match-data face)
"To MATCH-DATA apply FACE.
This is for use in buffers which have variable `font-lock-mode'
enabled; otherwise use `m-buffer-text-property-face'."
(m-buffer-put-text-property-match match-data
'font-lock-face face))
(provide 'm-buffer)
;;; m-buffer.el ends here
Next: m-buffer-macro, Previous: m-buffer, Up: Manipulate the Contents of Emacs Buffers [Contents]
m-buffer-at.el provides a set of stateless functions which for accessing data about buffers, without requiring changing the ‘current-buffer’.
Next: Commentary, Up: m-buffer-at [Contents]
This file is not part of Emacs
The contents of this file are subject to the GPL License, Version 3.0.
Copyright (C) 2014-2022 Free Software Foundation, Inc.
This program 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.
This program 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 this program. If not, see http://www.gnu.org/licenses/.
Next: Status, Previous: Header, Up: m-buffer-at [Contents]
Provides stateless equivalents to many core Emacs functions, that provide information about a buffer. Most of these functions take either a buffer as a parameter or a location, which is either a marker (with a non-nil buffer and location) or a buffer and integer.
These functions are generally competitive with the originals in terms of speed.
Next: Code, Previous: Commentary, Up: m-buffer-at [Contents]
There are lots more to do, but the interface should be stable.
Previous: Status, Up: m-buffer-at [Contents]
(require 'm-buffer-macro)
(defun m-buffer-at-point (buffer)
"Return the location of point in BUFFER.
See also `point'."
(with-current-buffer
buffer
(point)))
(defun m-buffer-at-eolp (&rest location)
"Return t if LOCATION is at the end of a line.
See also `eolp'."
(m-buffer-with-current-location
location
(eolp)))
(defun m-buffer-at-bolp (&rest location)
"Return t if LOCATION is at the begining of a line.
See also `bolp'"
(m-buffer-with-current-location
location
(bolp)))
(defun m-buffer-at-line-beginning-position (&rest location)
"Return the start of the line of LOCATION."
(m-buffer-with-current-location
location
(line-beginning-position)))
(defun m-buffer-at-line-end-position (&rest location)
"Return the end of the line of LOCATION."
(m-buffer-with-current-location
location
(line-end-position)))
(defun m-buffer-at-narrowed-p (buffer)
(with-current-buffer
buffer
(buffer-narrowed-p)))
(defun m-buffer-at-string (buffer)
(with-current-buffer
buffer
(buffer-string)))
(provide 'm-buffer-at)
;;; m-buffer-at.el ends here
Next: m-buffer-benchmark, Previous: m-buffer-at, Up: Manipulate the Contents of Emacs Buffers [Contents]
m-buffer-macro.el provides some general purpose macros for:
Next: Commentary, Up: m-buffer-macro [Contents]
This file is not part of Emacs
The contents of this file are subject to the GPL License, Version 3.0.
Copyright (C) 2014-2022 Free Software Foundation, Inc.
This program 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.
This program 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 this program. If not, see http://www.gnu.org/licenses/.
Next: Code, Previous: Header, Up: m-buffer-macro [Contents]
This file provides some utility macros which help to support stateless operation on buffers, by restoring global state after to what it was before the macro starts.
These macros are quite useful, but with the exception of ‘m-buffer-with-markers’, they are mostly meant to underpin ‘m-buffer-at’. The aim is that all the cases where one of these macros is used with a single form from core Emacs should be provided by m-buffer-at (although this is not the case yet). These macros might be more efficient if there are a lot of calls to group together.
Previous: Commentary, Up: m-buffer-macro [Contents]
Next: Point and Buffer, Up: Code [Contents]
Markers are generally much nicer than integers, but needs cleaning up afterwards if a lot are created. It’s possible to do this using ‘m-buffer-nil-marker’, but it can be a bit painful. This form looks like a ‘let’ form, but removes markers at the end.
(defmacro m-buffer-with-markers (varlist &rest body)
"Bind variables after VARLIST then eval BODY.
VARLIST is of the same form as `let'. All variables should
contain markers or collections of markers. All markers are niled
after BODY."
;; indent let part specially, and debug like let
(declare (indent 1)(debug let))
;; so, create a rtn var with make-symbol (for hygene)
(let* ((rtn-var (make-symbol "rtn-var"))
(marker-vars
(mapcar #'car varlist))
(full-varlist
(append
varlist
`((,rtn-var
(progn
,@body))))))
`(let* ,full-varlist
(m-buffer-nil-marker
(list ,@marker-vars))
,rtn-var)))
These macros are extensions of ‘with-current-buffer’, and ‘save-excursion’, which set the current buffer and location.
(defmacro m-buffer-with-current-marker
(marker &rest body)
"At MARKER location run BODY."
(declare (indent 1) (debug t))
`(with-current-buffer
(marker-buffer ,marker)
(save-excursion
(goto-char ,marker)
,@body)))
(defmacro m-buffer-with-current-position
(buffer location &rest body)
"In BUFFER at LOCATION, run BODY."
(declare (indent 2)
(debug t))
`(with-current-buffer
,buffer
(save-excursion
(goto-char ,location)
,@body)))
Combines the last two!
(defmacro m-buffer-with-current-location
(location &rest body)
"At LOCATION, run BODY.
LOCATION should be a list. If a one element list, it is a marker.
If a two element, it is a buffer and position."
(declare (indent 1) (debug t))
;; multiple eval of location!
(let ((loc (make-symbol "loc")))
`(let ((,loc ,location))
(if (= 1 (length ,loc))
(m-buffer-with-current-marker
(nth 0 ,loc)
,@body)
(if (= 2 (length ,loc))
(m-buffer-with-current-position
(nth 0 ,loc)
(nth 1 ,loc)
,@body)
(error "m-buffer-with-current-location requires a list of one or two elements"))))))
(provide 'm-buffer-macro)
;;; m-buffer-macro.el ends here
Next: Roadmap, Previous: m-buffer-macro, Up: Manipulate the Contents of Emacs Buffers [Contents]
m-buffer-benchmark.el provides no functions, but is a set of benchmarks to give some idea of how much overhead various m-buffer functions entail.
;;
Up: m-buffer-benchmark [Contents]
This file is not meant a emacs package, but for benchmarking m-buffer. To hide Emacs’ there are lots of places where m-buffer saves, changes and then restores this global state. One obvious question is what impact does this have on performance. We check this here.
Next: Support, Up: Benchmarking [Contents]
The results of running these forms are "pre-evaluated", because this file forms part of the lentic documentation for m-buffer. We could evaluate these at export time but, by default, this form of evaluation is blocked. Moreover, it can be quite slow which would be less than ideal with lentic-server.
To evaluate on the local machine use ‘org-babel-execute-buffer’, probably after setting ‘org-confirm-babel-evaluate’ to nil.
Next: How Long does it take to change current-buffer, Previous: Evaluation, Up: Benchmarking [Contents]
Build a nice simple bench macro.
(defmacro bench
(&rest body)
`(format "%e"
(car
(benchmark-run-compiled
1000000
(progn
,@body)))))
Previous: Support, Up: Benchmarking [Contents]
There are lots of places where we set the current buffer, do something, then get the result again, so understanding how long this takes is important. So, how long does it take to set and restore the current buffer.
It’s quite a bit slower – about an order of magnitude.
We get the ‘current-buffer’ and ‘point’. In the first case, we also do this inside a ‘with-current-buffer’.
(bench
(with-current-buffer
(current-buffer)
(point)))
(bench (current-buffer) (point))
Is ‘with-current-buffer’ quicker if we are already in the current-buffer? This is interesting to know because if it is, grouping several commands that operate on a single would run much faster.
We test this by entering having two ‘with-current-buffer’ calls which do nothing, one nested and one not. Our conclusion is, no, it makes not difference, so there is little pointing in putting a grouping construct in, unless we do something intelligent.
(bench
(with-current-buffer
(current-buffer)
(with-current-buffer
(current-buffer))))
(bench
(with-current-buffer
(current-buffer))
(with-current-buffer
(current-buffer)))
m-buffer-at provides stateless functions, but how much overhead does this introduce. We try this with the simplest function I can think of, which is point. The various forms look different here – because we have a ‘current-buffer’ call in, but not with ‘point’. But then, effectively, ‘point’ must call ‘current-buffer’ somewhere as part of its implementation, so this difference is fair.
We conclude that m-buffer is about 100x slower for calling ‘point’, even when the buffer does not actually need to be changed. So, a lot slower.
(bench (point))
(bench
(with-current-buffer
(current-buffer)
(point)))
(bench (m-buffer-at-point (current-buffer)))