aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorTom Ryder <tom@sanctum.geek.nz>2019-06-27 17:29:38 +1200
committerTom Ryder <tom@sanctum.geek.nz>2019-07-06 10:34:12 +1200
commit35d1c767d3e5f9ffdc52e100da2ce880d813554e (patch)
treed285f7cecf000bac44f4b679b562334398fc0227
parentRestore missing // sigils (diff)
downloaddotfiles-35d1c767d3e5f9ffdc52e100da2ce880d813554e.tar.gz
dotfiles-35d1c767d3e5f9ffdc52e100da2ce880d813554e.zip
Completely redo literate vimrc
-rw-r--r--vim/vimrc1387
1 files changed, 1279 insertions, 108 deletions
diff --git a/vim/vimrc b/vim/vimrc
index a8293311..1f5f1743 100644
--- a/vim/vimrc
+++ b/vim/vimrc
@@ -1,7 +1,8 @@
-" Tom Ryder (tejr)’s Literate Vimrc
-" =================================
+" ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+" Tom Ryder (tejr)’s Literate Vimrc
+" ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
"
-" Last updated: Thu, 27 Jun 2019 00:23:17 +0000
+" Last updated: Thu, 04 Jul 2019 10:28:57 +0000
"
" │ And I was lifted up in heart, and thought
" │ Of all my late-shown prowess in the lists,
@@ -13,10 +14,11 @@
" │
" │ —Tennyson
"
-" This is an attempt at something like a “literate vimrc”, in the tradition of
-" Donald Knuth’s “literate programming”: <http://www.literateprogramming.com/>
+" This file is an attempt at something like a “literate vimrc”, in the
+" tradition of Donald Knuth’s “literate programming”:
+" <http://www.literateprogramming.com/>
"
-" The dotfiles project containing this file is maintained here:
+" The dotfiles project that comprises it is maintained here:
" <https://sanctum.geek.nz/cgit/dotfiles.git>
"
" This is a long file, and comments abound within. Should this be bothersome,
@@ -26,33 +28,33 @@
" :g/\m^$\|^\s*"/d
"
" This file should be saved as ‘vimrc’—no leading period—in the user runtime
-" directory. On Unix-like operating systems, hereinafter referred to as
+" directory. On Unix-like operating systems, hereinafter referred to as
" “*nix”, that directory is ‘~/.vim’; on Windows, it’s ‘~/vimfiles’.
" Fortunately, those are the only two kinds of operating systems that exist,
" anywhere in the world.
"
-" It requires Vim 7.0 or newer, with the +eval feature, and the 'compatible'
-" option turned off, chiefly to allow line continuations. The vimrc stub at
-" ~/.vimrc (Unix) or ~/_vimrc (Windows) should check that these conditions are
-" met before loading this file with ‘:runtime vimrc’.
+" It requires Vim v7.0.0 or newer, with the +eval feature, and the
+" 'compatible' option turned off, chiefly to allow line continuations. The
+" vimrc stub at ~/.vimrc (Unix) or ~/_vimrc (Windows) should check that these
+" conditions are met before loading this file with ‘:runtime vimrc’.
"
" All of this should survive a pass of the Vim script linter Vint with no
" errors, warnings, or style problems: <https://github.com/Kuniwak/vint>
"
-" We’ll begin by making sure we’re all speaking the same language. Since it’s
-" been the future for a few years now, this file has characters outside the
-" ASCII character set, which prompts Vint to suggest declaring the file
-" encoding with a :scriptencoding command. The :help for that command
+" We’ll begin by making sure that we and Vim are speaking the same language.
+" Since it’s been the future for a few years now, this file has characters
+" outside the ASCII character set, which prompts Vint to suggest declaring the
+" file encoding with a :scriptencoding command. The :help for that command
" specifies that this should be done after 'encoding' is set, so we’ll do that
" here, too.
"
-" On *nix, I keep the primary locale environment variable $LANG defined, and
-" it almost always specifies a multibyte locale. This informs Vim’s choice of
-" internal character encoding, but the default for the 'encoding' option in
-" the absence of a valid $LANG is ‘latin1’. Since this is almost never what
-" I want, we’ll manually choose the UTF-8 encoding for Unicode in the absence
-" of any other explicit specification.
+" On *nix, I define the primary locale environment variable $LANG, almost
+" always specifying a multibyte locale. This informs Vim’s choice of internal
+" character encoding, but the default for the 'encoding' option in the absence
+" of a valid $LANG is ‘latin1’. Since this is almost never what I want, we’ll
+" manually choose the UTF-8 encoding for Unicode in the absence of any other
+" explicit specification.
"
if &encoding ==# 'latin1' && !exists('$LANG')
set encoding=utf-8
@@ -61,96 +63,276 @@ scriptencoding utf-8
" With encoding handled, we’ll turn our attention to the value of the
" 'runtimepath' option, since any scripts loaded from the paths specified
-" therein control so much of the behaviour of Vim.
-"
-" Working with 'runtimepath' and other options set with comma-separated
-" strings is error-prone, whether the strings are paths or not. In
-" particular, splitting the list is surprisingly complicated, as revealed by
-" the tokenizer function copy_option_part in src/misc2.c in Vim’s source code.
-" This awkwardness is largely because commas within each list item need to be
-" escaped with backslashes, but backslashes themselves are not. To make it
-" even more complicated, each separating comma may be followed by any number
-" of spaces, or more commas, that will be ignored. This means that you can
-" have a path starting with spaces or commas as the first value in the list,
-" but not any of the following values. Worse, there’s no way to escape
-" leading whitespace characters to prevent them being skipped; backslashes
-" don’t work, because they only escape commas in this context. Read the
-" source code if you don’t believe me. Vim, I love you, but you are really
-" weird.
-"
-" In an effort to abstract this away a bit, we’ll define a script-local
-" function that can split such values into their constituent parts. The
-" pattern required for the split() breaks down like this:
+" therein control so much of the behaviour of Vim. I’d like to do this as
+" accurately as possible, even with Vim’s unintuitive behavior around escaping
+" of these variables. One of the first things we’ll need to be able to do is
+" split the value of 'runtimepath' into its constituent path parts.
+"
+" Splitting the values of these comma-separated options correctly is
+" surprisingly complicated. The list separator for such options is more
+" accurately defined as follows:
+"
+" │ A comma not preceded by a backslash, and possibly followed by an arbitrary
+" │ number of spaces and commas.
+"
+" The pattern required for the split breaks down like this:
"
" \\ ← Literal backslash
" \@<! ← Negative lookbehind assertion; means that whatever occurred
-" before this pattern—in this case, a backslash—cannot precede what
+" before this pattern, here a backslash, cannot precede what
" follows, but anything that does precede it is not removed from
" the data as part of the split delimiter
" , ← Literal comma
" [, ]* ← Any number of commas and spaces
"
-" Once we have the individual elements, we have to remove the escaping for
-" periods, specifically remove up to one backslash before all periods. We do
-" that with a map() over substitute(), string-eval style, to accommodate older
-" Vim before Funcref variables were added. As explained above, we don’t need
-" to unescape backslashes themselves, as one might expect.
+" We don’t, however, have to deal with backslashes before other backslashes,
+" nor before any other character. You can read the source code for the ad-hoc
+" tokenizer in copy_option_part() in src/misc2.c in Vim’s source code, and
+" test it with some values of your own, if you want to understand why.
"
-function! s:OptionSplit(expr, ...) abort
- if a:0 > 1
- echoerr 'Too many arguments'
- endif
- let [expr, keepempty] = [a:expr, a:0 ? a:1 : 0]
- return map(
- \ split(expr, '\\\@<!,[, ]*', keepempty),
- \ 'substitute(v:val, ''\\,'', '','', ''g'')',
- \)
+" Vim, I love you, but you are really weird.
+"
+" Note that we’re calling a script-local wrapper around map() named Map(), and
+" making a function reference to a script-local function UnEscItem(), both of
+" which we’ll define shortly.
+"
+function! s:SplitOption(expr, ...) abort
+ let keepempty = a:0 ? a:1 : 0
+ let parts = split(a:expr, '\\\@<!,[, ]*', keepempty)
+ return s:Map(parts, function('s:UnEscItem'))
+endfunction
+
+" We declare a wrapper around map() to allow us always to call it with
+" a Funcref as the second function parameter, which isn’t directly supported
+" by map() until Vim v7.4.1989. If the running version is older than that,
+" apply string() to the Funcref to use the older calling convention.
+"
+" <https://github.com/vim/vim/releases/tag/v7.4.1989>
+"
+function! s:Map(list, Func) abort
+ return has('patch-7.4.1989')
+ \ ? map(a:list, a:Func)
+ \ : map(a:list, string(a:Func).'(0, v:val)')
+endfunction
+
+" We will need to be able to escape and unescape commas within separated list
+" items. As noted above, we do this by adding and removing a backslash before
+" each comma.
+"
+function! s:EscItem(item) abort
+ return escape(a:item, ',')
+endfunction
+function! s:UnEscItem(key, val) abort
+ return substitute(a:val, '\\,', ',', 'g')
endfunction
-" The next problem will be how we add values to the list correctly and safely.
-" We create a second script-local function that does the two rounds of
-" escaping necessary for a string that can be used in an :execute wrapper over
-" a :set command string.
-function! s:EscItemExec(item) abort
- let [item] = [a:item]
- return escape(escape(item, ','), '\ %#|"')
+" We will need a way to escape a string for general use in an :execute wrapper
+" to prevent it being interpreted as anything but a string. The fnameescape()
+" function, while somewhat misnamed, is the correct way to do this, but it
+" wasn’t added until Vim v7.1.299, so we’ll have to do our best to backport it
+" here.
+"
+" <https://github.com/vim/vim/releases/tag/v7.1.299>
+"
+function! s:EscArg(arg) abort
+ return exists('*fnameescape')
+ \ ? fnameescape(a:arg)
+ \ : escape(a:arg, "\n\r\t".' *?[{`$\%#''"|!<')
endfunction
-if exists('$MYVIM')
- execute 'set runtimepath^='.s:EscItemExec($MYVIM)
-elseif strlen(&runtimepath) > 0
- let s:runtimepath = s:OptionSplit(&runtimepath)
- let $MYVIM = s:runtimepath[0]
+
+" For the particular case of 'runtimepath', we also need to escape glob
+" characters like * to prevent them from being expanded.
+"
+function! s:EscWild(string) abort
+ let string = a:string
+ return escape(string, '\*?[{`''$~')
+endfunction
+
+" If an environment variable MYVIM exists, and it isn’t blank, apply its value
+" as the first value of 'runtimepath', after escaping it appropriately.
+" Otherwise, do it the other way around: the first path in the 'runtimepath'
+" list becomes MYVIM.
+"
+if exists('$MYVIM') && $MYVIM != ''
+ execute 'set runtimepath^='.s:EscArg(s:EscItem(s:EscWild(
+ \ $MYVIM
+ \)))
+elseif &runtimepath != ''
+ let $MYVIM = s:SplitOption(&runtimepath)[0]
endif
-function! s:Mkpath(path) abort
- let [path] = [a:path]
- return isdirectory(path)
- \ || exists('*mkdir') && mkdir(path)
+
+" We need a function to reliably create a full path, whether or not the
+" directories already exist. We create a wrapper with similar calling
+" conventions to mkdir(), but with the ‘p’ value for the second parameter
+" {prot} forced on. You can still provide alternative permissions in the
+" second argument.
+"
+function! s:CreatePath(name, ...) abort
+ if isdirectory(a:name)
+ return 1
+ endif
+ let prot = a:0 >= 1 ? a:1 : 0755
+ return mkdir(a:name, 'p', prot)
endfunction
-let s:cache = $MYVIM.'/cache'
-call s:Mkpath(s:cache)
-let &viminfo .= ',n'.s:cache.'/viminfo'
+
+" That’s a useful function, too, so we make it available to the user with
+" a user command. We’ll generally use the function form, as it requires less
+" escaping. An optional second argument can be provided, corresponding to the
+" mkdir() permissions parameter.
+"
+command! -bar -complete=dir -nargs=1 CreatePath
+ \ call s:CreatePath(<f-args>)
+
+" Now that we have a clean means to create directories if they don’t already
+" exist, let’s apply it for the first time to the user runtime directory.
+" Note that we aren’t checking whether this actually succeeded. We do want
+" errors raised if there were problems with the creation, but we’ll barrel on
+" ahead regardless after warning the user about our failure.
+"
+call s:CreatePath($MYVIM)
+
+" Our next application of our new :CreatePath command is to configure the path
+" for the viminfo metadata file, putting it in a cache subdirectory of the
+" user runtime directory set in MYVIM.
+"
+" Using this non-default location for viminfo has the nice benefit of
+" preventing command and search history from getting clobbered when something
+" runs Vim without using this vimrc, because such an instance will safely
+" write its history to the default viminfo path instead. It also contributes
+" to our aim of having everything related to the Vim runtime process in one
+" dedicated directory tree.
+"
+" The normal method of specifying the path to the viminfo file, as applied
+" here, is an addendum of the path to the 'viminfo' option with an "n" prefix.
+" Vim v8.1.716 introduced a nicer way to set this with an option named
+" 'viminfofile', but that’s still too new for us to use just yet.
+"
+" <https://github.com/vim/vim/releases/tag/v8.1.0716>
+"
+let s:viminfo = $MYVIM.'/viminfo'
+execute 'set viminfo+='.s:EscArg('n'.s:viminfo)
+
+" Speaking of recorded data in viminfo files, the default Vim limit of a mere
+" 50 entries for command and search history is pretty stingy. Because I don’t
+" think I’m ever likely to be in a situation where remembering several
+" thousand Vim commands and search patterns is going to severely tax memory,
+" let alone disk space, I’d rather this limit were much higher. It’s
+" sometimes really handy to dig up commands from many days ago.
+"
+" The maximum value for the 'history' option is documented in ‘:help
+" 'history'’ as 10000, so let’s just use that, and see if anything breaks.
+"
+set history=10000
+
+" We’ll now enable automatic backups of most file buffers, since that’s off by
+" default. In practice, I don’t need these backups very much, at least if I’m
+" using version control sensibly, but they have still saved my bacon a few
+" times.
+"
+" We’ll try to keep the backup files in a dedicated cache directory, to stop
+" them popping up next to the file to which they correspond, and getting
+" accidentally committed to version control.
+"
+" If Vim is new enough, we’ll add two trailing slashes to the path we’re
+" inserting, which prompts Vim to incorporate the full escaped path of the
+" relevant buffer in the backup filename, avoiding collisions.
+"
+" As a historical note, other similar directory path list options supported
+" this trailing slashes hint for a long time before 'backupdir' caught up to
+" them. The 'directory' option for swap files has supported it at least as
+" far back as v5.8.0 (2001), and 'undodir' appears to have supported it since
+" its creation in v7.2.438. Even though ‘:help 'backupdir'’ didn’t say so,
+" people assumed it would work the same way, when in fact Vim simply ignored
+" it until v8.1.0251.
+"
+" I don’t want to add the slashes to the option value in older versions of Vim
+" where they don’t do anything, so we’ll check the version ourselves to see if
+" there’s any point in including them.
+"
+" <https://github.com/vim/vim/releases/tag/v8.1.0251>
+"
+" It’s all so awkward. Surely separate options named something like
+" 'backupfullname', 'swapfilefullname' would have been clearer.
+"
set backup
-let s:backupdir = s:cache.'/backup'
-call s:Mkpath(s:backupdir)
-execute 'set backupdir^='.s:EscItemExec(
+let s:backupdir = $MYVIM.'/backup'
+call s:CreatePath(s:backupdir, 0700)
+execute 'set backupdir^='.s:EscArg(s:EscItem(
\ s:backupdir.(has('patch-8.1.251') ? '//' : ''),
- \)
+ \))
+
+" Files in certain directories on Unix-compatible filesystems should not be
+" backed up, for security reasons. This is particularly important if editing
+" temporary files created by sudoedit(8). On Unix-like systems, we here add
+" a few paths to the default value of 'backupskip' in order to prevent the
+" creation of such undesired backup files.
+"
+" * /dev/shm: RAM disk, default path for password-store’s temporary files
+" * /usr/tmp: Hard-coded path for sudoedit(8) [1/2]
+" * /var/tmp: Hard-coded path for sudoedit(8) [2/2]
+"
+" Prior to v8.1.1519, Vim didn’t check patterns added to 'backupskip' for
+" uniqueness, so adding the same path repeatedly resulted in duplicate strings
+" in the value. This was due to the absence of the P_NODUP flag for the
+" option’s definition in src/option.c in the Vim source code. If we’re using
+" a version older than v8.1.1519, we’ll need to explicitly reset 'backupskip'
+" to its default value before adding patterns to it, so that reloading this
+" file doesn’t stack up multiple copies of any added paths.
+"
+" <https://github.com/vim/vim/releases/tag/v8.1.1519>
+"
if has('unix')
if !has('patch-8.1.1519')
set backupskip&
endif
set backupskip^=/dev/shm/*,/usr/tmp/*,/var/tmp/*
endif
-let s:directory = s:cache.'/swap'
-call s:Mkpath(s:directory)
-execute 'set directory^='.s:EscItemExec(s:directory).'//'
+
+" Keep swap files for file buffers in a dedicated directory, rather than the
+" default of writing them to the same directory as the buffer file. Add two
+" trailing slashes to the path to prompt Vim to use the full escaped path in
+" its name, in order to avoid filename collisions, since the 'directory'
+" option has supported that hint for much longer than 'backupdir' has. We
+" apply CreatePath() to attempt to create the path first, if needed.
+"
+let s:directory = $MYVIM.'/swap'
+call s:CreatePath(s:directory, 0700)
+execute 'set directory^='.s:EscArg(s:EscItem(
+ \ s:directory.'//'
+ \))
+
+" Keep tracked undo history for files permanently, in a dedicated cache
+" directory, so that the u/:undo and CTRL-R/:redo commands will work between
+" Vim invocations.
+"
+" The 'undodir' option has the same structure as 'backupdir' and 'directory';
+" if we have a user runtime directory, create a sub-subdirectory within it
+" dedicated to the undo files cache. Note also the trailing double-slash as
+" a signal to Vim to use the full path of the original file in its undo file
+" cache’s name.
+"
+" Support for these persistent undo file caches was not released until v7.3.0,
+" so we need to check for the feature’s presence before we enable it.
+"
if has('persistent_undo')
set undofile
- let s:undodir = s:cache.'/undo'
- call s:Mkpath(s:undodir)
- execute 'set undodir^='.s:EscItemExec(s:undodir).'//'
+ let s:undodir = $MYVIM.'/undo'
+ call s:CreatePath(s:undodir, 0700)
+ execute 'set undodir^='.s:EscArg(s:EscItem(
+ \ s:undodir.'//'
+ \))
endif
+
+" Now that we have a bit more confidence in our runtime environment, set up
+" all of the filetype detection, plugin, and indent hooks.
+"
filetype plugin indent on
+
+" There are a couple of contexts in which it's useful to reload filetypes for
+" the current buffer, quietly doing nothing if filetypes aren't enabled.
+" We'll set up a user command named :ReloadFileType to do this, with
+" a script-local function backing it.
+"
function! s:ReloadFileType() abort
if exists('g:did_load_filetypes')
doautocmd filetypedetect BufRead
@@ -158,6 +340,30 @@ function! s:ReloadFileType() abort
endfunction
command! -bar ReloadFileType
\ call s:ReloadFileType()
+
+" We'll also define a :ReloadVimrc command. This may seem like overkill, at
+" first. Surely just `:source $MYVIMRC` would be good enough?
+"
+" The problem is there are potential side effects to the current buffer when
+" the vimrc is reloaded. The :set commands for options like 'expandtab' and
+" 'shiftwidth' may trample over different buffer-local settings that were
+" specified by filetype and indent plugins. To ensure these local values are
+" reinstated, we'll define the new command wrapper to issue a :ReloadFileType
+" command after the vimrc file is sourced.
+"
+" We can't put the actual :source command into the script-local function we
+" define here, because Vim would get upset that we're trying to redefine
+" a function as it executes!
+"
+" Just to be on the safe side, we also suppress any further ##SourceCmd hooks
+" from running the :source command with a :noautocmd wrapper. This is
+" a defensive measure to avoid infinite recursion. It may not actually be
+" necessary.
+"
+" We emit a message afterwards, just to make it clear that something has
+" happened. The :redraw just before that message seems to be necessary for
+" this message to display correctly. I'm not sure why.
+"
function! s:ReloadVimrc() abort
ReloadFileType
redraw
@@ -165,100 +371,599 @@ function! s:ReloadVimrc() abort
endfunction
command! -bar ReloadVimrc
\ noautocmd source $MYVIMRC | call s:ReloadVimrc()
+
+" We'll now create or reset a group of automatic command hooks specific to
+" matters related to reloading the vimrc itself, or maintaining and managing
+" options set within it.
+"
augroup vimrc
autocmd!
augroup END
+
+" Reload the stub vimrc, and thereby this main one, each time either of them
+" is written. This often makes errors in the file immediately apparent, and
+" saves restarting Vim or running the :source command manually, which I almost
+" always want to do after changing my vimrc file anyway.
+"
autocmd vimrc BufWritePost $MYVIMRC,$MYVIM/vimrc
\ ReloadVimrc
+
+" If Vim is new enough (v7.0.187) to support the ##SourceCmd event for
+" automatic command hooks, we'll also apply a hook for that event to catch
+" invocations of :source of either vimrc file, and translate that into
+" reloading the stub vimrc.
+"
+" <https://github.com/vim/vim/releases/tag/v7.0.187>
+"
if exists('##SourceCmd')
autocmd vimrc SourceCmd $MYVIMRC,$MYVIM/vimrc
\ ReloadVimrc
endif
-set history=10000
+
+" For spelling, use New Zealand English by default, but later on we’ll
+" configure a leader mapping to switch to United States English, since I so
+" often have to write for Yankees. We’ll set the 'spellfile' option too, to
+" place it in the cache directory into which we’ve been putting everything.
+" We’ll follow Vim’s standard naming convention for the file itself, though.
+" If available, my plugin spellfile_local.vim will extend this later to add
+" more spelling word lists per filetype and per file.
+"
+" We briefly set 'isfname' to every character but NUL if we’re using Unix,
+" since Vim uses it internally for 'spellfile' assignment to decide whether
+" the path is valid. We put it back immediately afterwards.
+"
set spelllang=en_nz
-let s:spellfile = s:cache.'/spell/'.join([
+let s:spelldir = $MYVIM.'/spell'
+call s:CreatePath(s:spelldir)
+let s:spellfile = s:spelldir.'/'.join([
\ split(&spelllang, '_')[0],
\ &encoding,
\ 'add',
\], '.')
-execute 'set spellfile='.s:EscItemExec(s:spellfile)
+if has('unix')
+ let s:isfname = &isfname
+ set isfname=1-255
+endif
+execute 'set spellfile^='.s:EscArg(s:EscItem(s:spellfile))
+let &isfname = s:isfname
+
+" Spell checking includes optional support for catching lower case letters at
+" the start of sentences, and defines a pattern in 'spellcapcheck' for the end
+" of a sentence. The default is pretty good, but with two-spacing with
+" 'cpoptions' including ‘J’ and 'formatoptions' including ‘p’ as set later in
+" this file, we can be less ambiguous in this pattern. We require two
+" consecutive spaces, a newline, a carriage return, or a tab to mark the end
+" of a sentence. This means that we could make abbreviations like "i.e.
+" something" without flagging "something" as a spelling error.
+"
+" We use :let syntax rather than :set here, just to avoid a whole bunch of
+" annoying escaping for the value.
+"
let &spellcapcheck = '[.?!]\%( \|[\n\r\t]\)'
+
+" For word completion in insert mode with CTRL-X CTRL-K, or if 'complete'
+" includes the ‘k’ flag, the 'dictionary' option specifies the path to the
+" system word list. This makes the dictionary completion work consistently,
+" even if 'spell' isn’t set at the time to coax it into using 'spellfile'.
+"
+" It’s not an error if the system directory file added first doesn’t exist;
+" it’s just a common location that often yields a workable word list, and does
+" so on all of my main machines.
+"
+" At some point, I may end up having to set this option along with 'spellfile'
+" a bit more intelligently to ensure that spell checking and dictionary
+" function consistently, and with reference to the same resources. For the
+" moment, I’ve just added another entry referring to a directory in the user
+" runtime directory, but I don’t have anything distinct to put there yet.
+"
+" In much the same way, we add an expected path to a thesaurus, for completion
+" with CTRL-X CTRL-T in insert mode, or with ‘t’ added to 'completeopt'. The
+" thesaurus data isn’t installed as part of the default ‘install-vim’ target
+" in tejr’s dotfiles, but it can be retrieved and installed with
+" ‘install-vim-thesaurus’.
+"
+" I got the thesaurus itself from the link in the :help for 'thesaurus' in
+" v8.1. It’s from WordNet and MyThes-1. I maintain a mirror on my own
+" website that the Makefile recipe attempts to retrieve. I had to remove the
+" first two metadata lines from thesaurus.txt, as Vim appeared to interpret
+" them as part of the body data.
+"
+" Extra checks for appending the 'dictionary' and 'thesaurus' paths in MYVIM
+" need to be made, because the P_NDNAME property is assigned to them, which
+" enforces a character blacklist in the option value. We check for the
+" expected Vim error code here, and if the MYVIM path offends, we just skip
+" the setting entirely, rather than throwing cryptic errors at the user. None
+" of the blacklisted characters are particularly wise characters to have in
+" paths, anyway, legal though they may be on Unix filesystems. We can’t work
+" around this one with 'isfname'; the blacklist is hardcoded.
+"
set dictionary^=/usr/share/dict/words
let s:ref = $MYVIM.'/ref'
-let s:dictionary = s:ref.'/dictionary.txt'
-execute 'set dictionary^='.s:EscItemExec(s:dictionary)
-let s:thesaurus = s:ref.'/thesaurus.txt'
-execute 'set thesaurus^='.s:EscItemExec(s:thesaurus)
+try
+ execute 'set dictionary^='.s:EscArg(s:EscItem(
+ \ s:ref.'/dictionary.txt'
+ \))
+ execute 'set thesaurus^='.s:EscArg(s:EscItem(
+ \ s:ref.'/thesaurus.txt'
+ \))
+catch /^Vim\%((\a\+)\)\=:E474:/
+endtry
+
+" Next, we’ll modernize a little in adjusting some options with old
+" language-specific defaults.
+"
+" Traditional vi was often used for development in the C programming language.
+" The default values for a lot of Vim’s options still reflect this common use
+" pattern. In this case, the 'comments' and 'commentstring' options reflect
+" the C syntax for comments:
+"
+" /*
+" * This is an ANSI C comment.
+" */
+"
+" Similarly, the 'define' and 'include' options default to C preprocessor
+" directives:
+"
+" #define FOO "bar"
+"
+" #include "baz.h"
+"
+" Times change, however, and I don’t get to work with C nearly as much as I’d
+" like. The defaults for these options no longer make sense, and so we blank
+" them, compelling filetype plugins to set them as they need instead.
+"
+" The default value for the 'path' option is similar, in that it has an aged
+" default; this option specifies directories in which project files and
+" includes can be unearthed by navigation commands like 'gf'. Specifically,
+" its default value comprises /usr/include, which is another C default. Let’s
+" get rid of that, too.
+"
set comments= commentstring= define= include=
set path-=/usr/include
+
+" Next, we’ll adjust the global indentation settings. In general and as
+" a default, I prefer spaces to tabs, and I like to use four of them, for
+" a more distinct visual structure. Should you happen to disagree with this,
+" I cordially invite you to fite me irl.
+"
+" <https://sanctum.geek.nz/blinkenlights/spaces.webm>
+"
+" Filetype indent plugins will often refine these settings for individual
+" buffers. For example, 'expandtab' is not appropriate for Makefiles, nor for
+" the Go programming language. For another, two-space indents are more
+" traditional for Vim script.
+"
set autoindent
set expandtab
set shiftwidth=4
set smarttab
+
+" Apply 'softtabstop' option to make a tab key press in insert mode insert the
+" same number of spaces as defined by the indent depth in 'shiftwidth'. If
+" Vim is new enough to support it (v7.3.693), apply a negative value to do
+" this dynamically if 'shiftwidth' changes.
+"
+" <https://github.com/vim/vim/releases/tag/v7.3.693>
+"
if v:version > 703 || v:version == 703 && has('patch693')
set softtabstop=-1
else
let &softtabstop = &shiftwidth
endif
-set backspace+=eol
-set backspace+=indent
-set backspace+=start
+
+" Relax traditional vi’s harsh standards over what regions of the buffer can
+" be removed with backspace in insert mode. While this admittedly allows bad
+" habits to continue, since insert mode by definition is not really intended
+" for deleting text, I feel the convenience outweighs that in this case.
+"
+set backspace+=eol " Line breaks
+set backspace+=indent " Leading whitespace characters created by 'autoindent'
+set backspace+=start " Text before the start of the current insertion
+
+" When soft-wrapping text with the 'wrap' option on, which is off by default,
+" break the lines between words, rather than within them; it’s much easier to
+" read.
+"
set linebreak
+
+" Similarly, show that the screen line is a trailing part of a wrapped line by
+" prefixing it with an ellipsis. If we have a multi-byte encoding, use
+" a proper ellipsis character to save a couple of columns, but otherwise three
+" periods will do just fine.
+"
+" … U+2026 HORIZONTAL ELLIPSIS
+"
+" Note that we test for the presence of a multi-byte encoding with a special
+" feature from ‘:help feature-list’, as recommended by ‘:help encoding’.
+" Checking that ‘&encoding ==# 'utf-8'’ is not quite the same thing, though
+" it’s unlikely I’ll ever use a different Unicode encoding by choice.
+"
if has('multi_byte_encoding')
set showbreak=…
else
set showbreak=...
endif
+
+" The visual structure of code provided by indents breaks down if a lot of the
+" lines wrap. Ideally, most if not all lines would be kept below 80
+" characters, but in cases where this isn’t possible, soft-wrapping longer
+" lines when 'wrap' is on so that the indent is preserved in the following
+" line mitigates this breakdown somewhat.
+"
+" With this 'breakindent' option set, it’s particularly important to have
+" 'showbreak' set to something besides an empty string, as done above,
+" otherwise without line numbers it’s hard to tell what’s a logical line and
+" what’s not.
+"
+" The 'breakindent' option wasn’t added until v7.4.338, so we need to check it
+" exists before we set it.
+"
+" <https://github.com/vim/vim/releases/tag/v7.4.338>
+"
if exists('+breakindent')
set breakindent
endif
+
+" Rather than rejecting operations like :write or :saveas when 'readonly' is
+" set or in other situations in which data might be lost, Vim should give me
+" a prompt to allow me to confirm that I know what I’m doing.
+"
set confirm
+
+" If Vim receives an Escape key code in insert mode, it shouldn’t wait to see
+" if it’s going to be followed by another key code, despite this being how the
+" function keys and Meta/Alt modifier are implemented for many terminal types.
+" Otherwise, if I press Escape, there’s an annoying delay before 'showmode'
+" stops showing '--INSERT--'.
+"
+" This breaks the function keys and the Meta/Alt modifier in insert mode in
+" most or maybe all of the terminals I use, but I don’t want those keys in
+" insert mode, anyway. All of this works fine in the GUI, of course.
+"
set noesckeys
+
+" Automatic text wrapping options using flags in the 'formatoptions' option
+" begin here. I rely on the filetype plugins to set the ‘t’ and ‘c’ flags for
+" this option to configure whether text or comments should be wrapped, as
+" appropriate for the document type or language, and so I don’t mess with
+" either of those flags here.
+
+" If a line is already longer than 'textwidth' would otherwise limit when
+" editing of that line begins in insert mode, don’t suddenly automatically
+" wrap it; I’ll break it apart myself with a command like 'gq'. This doesn’t
+" seem to stop paragraph reformatting with ‘a’, if that’s set.
+"
set formatoptions+=l
+
+" Don’t wrap a line in such a way that a single-letter word like "I" or "a" is
+" at the end of it. Typographically, as far as I can tell, this seems to be
+" a stylistic preference rather than a rule, rather like avoiding "widow" and
+" "orphan" lines in typesetting. I think it generally looks better to have
+" the short word start the line, so we’ll switch it on.
+"
set formatoptions+=1
+
+" If the filetype plugins have correctly described what the comment syntax for
+" the buffer’s language looks like, it makes sense to use that to figure out
+" how to join lines within comments without redundant comment syntax cropping
+" up. For example, with this set, joining lines in this very comment with ‘J’
+" would remove the leading ‘"’ characters.
+"
+" This 'formatoptions' flag wasn’t added until v7.3.541. Because we can’t
+" test for the availability of option flags directly, we resort to a version
+" number check before attempting to set it. I don’t like using :silent! to
+" suppress errors for this sort of thing when I can reasonably avoid it, even
+" if the tests are somewhat more verbose.
+"
+" <https://github.com/vim/vim/releases/tag/v7.3.541>
+"
if v:version > 703 || v:version == 703 && has('patch541')
set formatoptions+=j
endif
+
+" A momentary digression here into the doldrums of 'cpoptions'—after staunchly
+" opposing it for years, I have converted to two-spacing. You can blame Steve
+" Losh:
+"
+" <http://stevelosh.com/blog/2012/10/why-i-two-space/>
+"
+" Consequently, we specify that sentence objects for the purposes of the ‘s’
+" text object, the ‘(’ and ‘)’ sentence motions, and formatting with the 'gq'
+" command must be separated by *two* spaces. One space does not suffice.
+"
+" My defection to the two-spacers is also the reason I now leave 'joinspaces'
+" set, per its default, so that two spaces are inserted when consecutive
+" sentences separated by a line break are joined onto one line by the ‘J’
+" command.
+"
set cpoptions+=J
+
+" Separating sentences with two spaces has an advantage in making a clear
+" distinction between two different types of periods: periods that abbreviate
+" longer words, as in “Mr. Moolenaar”, and periods that terminate sentences,
+" like this one.
+"
+" If we’re using two-period spacing for sentences, Vim can interpret the
+" different spacing to distinguish between the two types, and can thereby
+" avoid breaking a line just after an abbreviating period. For example, the
+" two words in “Mr. Moolenaar” should never be split apart, preventing
+" confusion on the reader’s part lest the word "Mr." look too much like the
+" end of a sentence, and also preserving the semantics of that same period for
+" subsequent reformatting; its single-space won’t get lost.
+"
+" So, getting back to our 'formatoptions' settings, that is what the ‘p’ flag
+" does. I wrote the patch that added it, after becoming envious of an
+" analogous feature during an ill-fated foray into GNU Emacs usage.
+"
+" <https://github.com/vim/vim/releases/tag/v8.1.1523>
+"
if has('patch-8.1.728')
set formatoptions+=p
endif
+
+" In an effort to avoid loading unnecessary files, we add a flag to the
+" 'guioptions' option to prevent the menu.vim runtime file from being loaded.
+" It doesn’t do any harm, but I never use it, and it’s easy to turn it off.
+"
+" The documentation for this flag in ‘:help 'go-M'’ includes a note saying the
+" flag should be set here, rather that in the GUI-specific gvimrc file, as one
+" might otherwise think.
+"
if has('gui_running')
set guioptions+=M
endif
+
+" By default, Vim doesn’t allow a file buffer to have unwritten changes if
+" it’s not displayed in a window. Setting this option removes that
+" restriction so that buffers can remain in a modified state while not
+" actually displayed anywhere.
+"
+" This option is set in almost every vimrc I read; it’s so pervasive that
+" I sometimes see comments expressing astonishment or annoyance that it isn’t
+" set by default. However, I didn’t actually need this option for several
+" years of Vim usage, because I instinctively close windows onto buffers only
+" after the buffers within them have been written anyway.
+"
+" However, the option really is required for batch operations performed with
+" commands like :argdo or :bufdo, because Vim won’t otherwise tolerate
+" unwritten changes to a litany of buffers that are not displayed in any
+" window. After I started using such command maps a bit more often,
+" I realized I finally had a reason to turn this on permanently.
+"
set hidden
+
+" Highlight matches for completed searches in the buffer text, but clear that
+" highlighting away when this vimrc file is reloaded. Later on in this file,
+" CTRL-L in normal mode is remapped to issue :nohlsearch in addition to its
+" usual screen refresh function.
+"
set hlsearch
nohlsearch
+
+" Highlight search matches in my text while I’m still typing my pattern,
+" including scrolling the screen to show the first such match if necessary.
+" This can be somewhat jarring, particularly when the cursor ends up scrolling
+" a long way from home in a large file, but I think the benefits of being able
+" to see instances of what I’m trying to match as I type the pattern do
+" outweigh that discomfort.
+"
set incsearch
+
+" Don’t waste cycles and bandwidth redrawing the screen during execution of
+" macro recordings and scripts.
+"
set lazyredraw
-set listchars+=tab:>-
-set listchars+=trail:-
-set listchars+=nbsp:+
+
+" Define meta-characters to show in place of characters that are otherwise
+" invisible, or line wrapping attributes when the 'list' option is enabled.
+"
+" These 'list' characters all correspond to invisible or indistinguishable
+" characters. We leave the default eol:$ in place to show newlines, and add
+" a few more.
+"
+set listchars+=tab:>- " Tab characters, preserve width with hyphens
+set listchars+=trail:- " Trailing spaces
+set listchars+=nbsp:+ " Non-breaking spaces
+
+" The next pair of 'list' characters are arguably somewhat misplaced, in that
+" they don’t really represent invisible characters in the same way as the
+" others, but are hints for the presence of other characters on unwrapped
+" lines that are wider than the screen. They’re very useful, though.
+"
+" If the current encoding supports it, use these non-ASCII characters for the
+" markers, as they’re visually distinctive:
+"
+" extends: Signals presence of unwrapped text to screen right
+" » U+00BB RIGHT-POINTING DOUBLE ANGLE QUOTATION MARK
+" precedes: Signals presence of unwrapped text to screen left
+" « U+00BB LEFT-POINTING DOUBLE ANGLE QUOTATION MARK
+"
+" Failing that, ‘<’ and ‘>’ will do the trick.
+"
if has('multi_byte_encoding')
set listchars+=extends:»,precedes:«
else
set listchars+=extends:>,precedes:<
endif
+
+" Don’t let your editor’s options be configured by content in arbitrary files!
+" Down with modelines! Purge them from your files! Écrasez l’infâme!
+"
+" I think that modelines are Vim’s worst misfeature, and that 'nomodeline'
+" should be the default. It’s enabled pretty bad security vulnerabilities
+" over the years, and it’s a lot more effective to use filetype detection,
+" other automatic command hooks, or methods like .editorconfig to set
+" variables specifically for a buffer or project.
+"
set nomodeline
+
+" The only octal numbers I can think of that I ever even encounter are Unix
+" permissions masks, and I’d never use CTRL-A or CTRL-X to increment them.
+" Numbers with leading zeroes are far more likely to be decimals.
+"
set nrformats-=octal
+
+" I like to leave the last line of the screen blank unless something is
+" actually happening in the editor for it to report, so I have grown to like
+" the Vim default of 'noruler'. CTRL-G shows me everything I need to know,
+" and is near-instinctive now.
+"
+" Rude system vimrc files tend to switch this back on, though, so we force it
+" off here.
+"
set noruler
-set sessionoptions-=localoptions
-set sessionoptions-=options
+
+" Sessions preserve window, tab, and buffer layout, and are thereby great for
+" more complex and longer-term projects like books, but they don’t play
+" together well with plugins and filetype plugins. Restoring the same
+" settings from both reloaded plugins and from the session causes screeds of
+" errors. Adjusting session behavior to stop it trying to restore the sorts
+" of settings that plugins manage makes them usable again.
+"
+set sessionoptions-=localoptions " No buffer options or mappings
+set sessionoptions-=options " No global options or mappings
+
+" Turn 'showcmd' off if a system vimrc has been rude enough to set it; I don’t
+" like how it can interfere with the display of longer lines.
+"
set noshowcmd
+
+" The ‘I’ flag for the 'shortmess' option prevents the display of the Vim
+" startup screen with version information, :help hints, and donation
+" suggestion. After I registered Vim and donated to Uganda per the screen’s
+" plea, I didn’t feel bad about turning this off anymore. Even with this
+" setting in place, I wouldn’t normally see it too often anyway, as I seldom
+" start Vim with no file arguments.
+"
+" I haven’t felt the need to mess with the other flags in this option.
+" I don’t have any problems with spurious Enter prompts, which seems to be the
+" main reason people pile it full of letters.
+"
set shortmess+=I
-if !&loadplugins || globpath(&runtimepath, 'plugin/matchparen.vim') ==# ''
+
+" We declare a function just to make a slightly more readable way to express
+" a check that plugins are going to be loaded and that a plugin of a given
+" name appears to be available somewhere in one of the runtime paths.
+"
+function! s:PluginReady(name) abort
+ return &loadplugins
+ \ && globpath(&runtimepath, 'plugin/'.a:name.'.vim') != ''
+endfunction
+
+" We’ll only use the old 'showmatch' method of a brief jump to the matching
+" bracket under the cursor if the much-preferred matchparen.vim standard
+" plugin doesn’t look like it’s going to load, whether because plugins have
+" been disabled, or it’s not in any of the plugin directories.
+"
+if !s:PluginReady('matchparen')
set showmatch matchtime=3
endif
+
+" I find the defaults of new windows opening above or to the left of the
+" previous window too jarring, because I’m used to both the i3 window manager
+" and the tmux terminal multiplexer doing it the other way around, in reading
+" order. I prefer the visual effect of the previous text staying where it is,
+" and the new window occupying previously blank space.
+"
set splitbelow splitright
+
+" Limit the number of characters per line that syntax highlighting will
+" attempt to match. This is as much an effort to encourage me to break long
+" lines and do hard wrapping correctly as it is for efficiency.
+"
set synmaxcol=500
+
+" Vim has an internal list of terminal types that support using smoother
+" terminal redrawing, and for which 'ttyfast' is normally set, described in
+" ‘:help 'ttyfast'’. That list includes most of the terminals I use, but
+" there are a couple more for which the 'ttyfast' option should apply: the
+" windows terminal emulator PuTTY, and the terminal multiplexer tmux, both of
+" which I use heavily.
+"
if &term =~# '^putty\|^tmux'
set ttyfast
endif
+
+" We really don’t want a mouse; while I use it a lot for cut and paste in X,
+" it just gets in the way if the tool running in the terminal tries to use it
+" too. Mouse events should be exclusively handled by the terminal emulator
+" application, so Vim shouldn’t try to give me terminal mouse support, even if
+" it would work.
+"
+" The manual suggests that disabling this should be done by clearing 't_RV',
+" but that didn’t actually seem to work when I tried it.
+"
set ttymouse=
+
+" While using virtual block mode, allow me to navigate to any column of the
+" buffer window; don’t confine the boundaries of the block to the coordinates
+" of characters that actually exist in the buffer text. While working with
+" formatted columnar data with this off is generally OK, it’s a hassle for
+" more subtle applications of visual block mode.
+"
set virtualedit+=block
+
+" I can’t recall a time that Vim’s error beeping or flashing was actually
+" useful to me, and so we turn it off in the manner that the manual instructs
+" in ‘:help 'visualbell'’. This enables visual rather than audio error bells,
+" but in the same breath, blanks the terminal attribute that would be used to
+" trigger such screen blinking, indirectly disabling the bell altogether.
+"
+" I thought at first that the newer 'belloff' and/or 'errorbells' options
+" would be a more intuitive way to keep Vim quiet, but the last time I checked
+" that, neither appeared to work as comprehensively as this older method does.
+"
+" Interestingly, the :help says that this setting has to be repeated in the
+" gvimrc file for GUI Vim.
+"
set visualbell t_vb=
+
+" When Ex command line completion is started with Tab, list valid completions
+" and complete the command line to the longest common substring, just as Bash
+" does, with just the one key press.
+"
+" The default value of 'full' for the 'wildmode' option puts the full
+" completion onto the line immediately, which I tolerate for insert mode
+" completion but don’t really like on the Ex command line. Instead, I arrange
+" for that to happen only with a second key press.
+"
set wildmenu
set wildmode=list:longest,full
+
+" Define a list of patterns to ignore for file and directory command line
+" completion. Files and directories with names matching any of these patterns
+" won’t be presented as candidates for tab completion on the command line.
+"
+" To make this list, I went right through my home directory with
+" a ‘find’-toothed comb; counted the occurrences of every extension, forced
+" down to lowercase; and then manually selected the ones that I was confident
+" would seldom contain plain text.
+"
+" The following incantation does the trick with POSIX-compatible shell tools,
+" giving patterns for the top 100 alphanumeric extensions for files from the
+" running user’s home directory:
+"
+" $ (LC_ALL=C find "$HOME" ! -type d -name '*.?*' -exec \
+" sh -c 'for fn ; do
+" ext=${fn##*.}
+" case $ext in
+" (*[![:alnum:]]*) continue ;;
+" (?*) printf "%s\n" "$ext" ;;
+" esac
+" done' _ {} + |
+" tr '[[:upper:]]' '[[:lower:]]' | sort | uniq -c |
+" sort -k1,1nr | awk 'NR <= 100 {print "*." $2}')
+"
+" I turned out to have rather a lot of .html and .vim files.
+"
+" If you’re scoffing at that and thinking "I could write a much simpler one",
+" please do so, and send it to me at <tom@sanctum.geek.nz> to have yours put
+" in here instead, with appropriate credit. Don’t forget to handle more than
+" ARG_MAX files, include filenames with newlines, and that the -z or -0 null
+" separator extensions are not standardized in POSIX.
+"
+" <https://mywiki.wooledge.org/UsingFind#Complex_actions>
+"
set wildignore=*~,#*#
\,*.7z
\,.DS_Store
@@ -330,213 +1035,679 @@ set wildignore=*~,#*#
\,*.xpm
\,*.xz
\,*.zip
+
+" Allow me to be lazy and type a path to complete on the Ex command line in
+" all-lowercase, and transform the consequent completion to match the
+" appropriate case, like the Readline setting completion-ignore-case can be
+" used for GNU Bash.
+"
+" As far as I can tell, despite its name, the 'wildignore' case option doesn’t
+" have anything to do with the 'wildignore' option, and so files that would
+" match any of those patterns only with case insensitivity implied will still
+" be candidates for completion.
+"
+" The option wasn’t added until v7.3.72, so we need to check it exists before
+" we try to set it.
+"
+" <https://github.com/vim/vim/releases/tag/v7.3.072>
+"
if exists('+wildignorecase')
set wildignorecase
endif
+
+" Enable syntax highlighting, but only if it’s not already on, to save
+" reloading the syntax files unnecessarily.
+"
+" <https://sanctum.geek.nz/blinkenlights/syntax-on.jpg>
+"
+" For several months in 2018, as an experiment, I tried using terminals with
+" no color at all, imitating a phenomenally productive BSD purist co-worker
+" who abhorred color in any form on his terminals. He only drank black
+" coffee, too. If you’re reading this: Hello, bdh!
+"
+" That experiment was instructive and interesting, and I found I had been
+" leaning on color information in some surprising ways. However, some months
+" later, I found I still missed my colors, and so I went back to my
+" Kodachrome roots, and didn’t pine at all for that monochrome world.
+"
+" The thing I most like about syntax highlighting is detecting runaway
+" strings, which generally works in even the most threadbare language syntax
+" highlighting definitions. I kept missing such errors when I didn’t have the
+" colors. I don’t have high standards for it otherwise, except maybe for
+" shell script.
+"
if !exists('syntax_on')
syntax enable
endif
+
+" Before we attempt to pick a syntax highlighting color scheme, we’ll set up
+" a couple of hooks for color scheme loading. In this case, we turn
+" 'cursorline' on if my 'sahara' color scheme is loaded, since I’ve configured
+" it to be a very dark gray that doesn’t stand out too much against a black
+" background. For any other color scheme, turn the option off, because it
+" almost always stands out too much for my liking.
+"
+" You’d think the pattern here could be used to match the color scheme name,
+" and it can be...after patch v7.4.108, when Christian Brabandt fixed it.
+" Until that version, it matched against the current buffer name, so we’re
+" forced to have an explicit test in the command instead.
+"
+" <https://github.com/vim/vim/releases/tag/v7.4.108>
+"
autocmd vimrc ColorScheme *
\ let &cursorline = g:colors_name ==# 'sahara'
+
+" Use 'dark' as my default value for 'background', in the absence of an
+" environment variable COLORFGBG or a response in v:termrbgresp that would set
+" it specifically.
+"
if !exists('$COLORFGBG') && get(v:, 'termrbgresp', '') ==# ''
set background=dark
endif
+
+" If the background seems to be dark, and I have either the GUI or a 256 color
+" terminal, and my custom sahara.vim color scheme looks to be available, load
+" it.
+"
if &background ==# 'dark'
\ && (has('gui_running') || &t_Co >= 256)
\ && globpath(&runtimepath, 'colors/sahara.vim') !=# ''
colorscheme sahara
endif
+
+" My mapping definitions begin here. I have some general personal rules for
+" approaches to mappings:
+"
+" * Use the configured Leader key as a prefix for mappings as much as
+" possible.
+"
+" * Use only the configured LocalLeader key as a prefix for mappings that are
+" defined as local to a buffer, which for me are almost always based on
+" &filetype and set up by ftplugin files.
+"
+" * If a normal mode map would make sense in visual mode, take the time to
+" configure that too. Use :xmap and its analogues rather than :vmap to
+" avoid defining unusable select-mode mappings, even though I never actually
+" use selection mode directly.
+"
+" * Avoid mapping in insert mode; let characters be literal to the greatest
+" extent possible, and avoid "doing more" in insert mode besides merely
+" inserting text as it’s typed.
+"
+" * Avoid key chords with Ctrl in favor of leader keys.
+"
+" * Never use Alt/Meta key chords; the terminal support for them is just too
+" confusing and flaky.
+"
+" * Don’t suppress display of mapped commands for no reason; it’s OK to show
+" the user the command that’s being run under the hood. Do avoid HIT-ENTER
+" prompts, though.
+"
+" * Avoid shadowing any of Vim’s existing functionality. If possible, extend
+" or supplement what Vim does, rather than replacing it.
+"
+" We’ll start with the non-leader mappings. Ideally, there shouldn’t be too
+" many of these.
+"
+
+" Use backspace as an even quicker way to switch to the current buffer’s
+" alternate buffer. User nickspoons of #vim was incredulous that I had never
+" used CTRL-^ and indeed did not know about it. I have since repented.
+"
nnoremap <Backspace> <C-^>
+
+" I find the space bar’s default behavior in normal mode of moving right one
+" character to be useless. Instead, I remap it to be a lazy way of paging
+" through the argument list buffers, scrolling a page until the last line of
+" the buffer is visible, and then moving to the :next buffer.
+"
+" I always wanted you to go into space, man.
+"
nnoremap <expr> <Space>
\ line('w$') < line('$')
\ ? "\<PageDown>"
\ : ":\<C-U>next\<CR>"
-if &loadplugins && globpath(&runtimepath, 'plugin/insert_cancel.vim') !=# ''
+
+" I hate CTRL-C’s default insert mode behavior. It ends the insert session
+" without firing the InsertLeave event for automatic command hooks. Why would
+" anyone want that? It breaks plugins that hinge on mirrored functionality
+" between the InsertEnter and InsertLeave events, and doesn’t otherwise differ
+" from Escape or :stopinsert. Even worse, people think it’s a *synonym* for
+" Escape, and use it because it’s easier to reach than the Escape key or
+" CTRL-[. Terrible!
+"
+" Instead, I apply a custom plugin named insert_cancel.vim to make it cancel
+" the current insert operation; that is, if the buffer has changed at all
+" since the start of the insert operation, pressing CTRL-C will reverse it,
+" while ending insert mode and firing InsertLeave as normal. This makes way
+" more sense to me, and I use it all the time now.
+"
+" <https://sanctum.geek.nz/cgit/vim-insert-cancel.git/about/>
+"
+" You might think on a first look, as I did, that a plugin is overkill, and
+" that a mapping like this would be all that’s required:
+"
+" :inoremap <C-C> <Esc>u
+"
+" Indeed, it *mostly* works, but there are some subtle problems with it. The
+" primary issue is that if you didn’t make any changes during the insert mode
+" session that you’re terminating, it *still* reverses the previous change,
+" which will be something else entirely that you probably *didn’t* mean to be
+" undone. The plugin’s way of working around this and the other shortcomings
+" of the simple mapping above is not too much more complicated, but it was not
+" easy to figure out.
+"
+" At any rate, we only want to establish the mapping if we can expect the
+" plugin to load, so test that 'loadplugins' is set and that the plugin file
+" exists with the expected filename.
+"
+" If the plugin isn’t available, I just abandon CTRL-C to continue its
+" uselessness.
+"
+if s:PluginReady('insert_cancel')
imap <C-C> <Plug>(InsertCancel)
endif
+
+" I often can’t remember (or guess) digraph codes, and want to look up how to
+" compose a specific character that I can name, at least in part. The table
+" in ‘:help digraph-table’ is what to use for that situation, and it solves
+" the problem, but I didn’t like the overhead of repeated lookups therein.
+"
+" Steve Losh has a solution I liked where a double-tap of CTRL-K in insert
+" mode brought up the applicable :help window:
+"
+" <https://bitbucket.org/sjl/dotfiles/src/2559256/vim/vimrc#lines-309:310>
+"
+" I took that one step further with a custom plugin named digraph_search.vim.
+" It parses the digraph table from :help and runs a simple text search of its
+" names using a string provided by the user. For example, searching for ACUTE
+" yields:
+"
+" > Digraphs matching ACUTE:
+" > ´ '' ACUTE ACCENT
+" > Á A' LATIN CAPITAL LETTER A WITH ACUTE
+" > É E' LATIN CAPITAL LETTER E WITH ACUTE
+" > Í I' LATIN CAPITAL LETTER I WITH ACUTE
+" > ...etc...
+"
+" <https://sanctum.geek.nz/cgit/vim-digraph-search.git/about/>
+"
+" This leaves you in insert mode, ready to hit CTRL-K one more time and then
+" type the digraph that you’ve hopefully found.
+"
+" Since a double-tap of CTRL-K does nothing in default Vim, we don’t bother
+" checking that the plugin’s available before we map to it; it’ll just quietly
+" do nothing.
+"
imap <C-K><C-K> <Plug>(DigraphSearch)
+
+" I end up hitting CTRL-L to clear or redraw the screen in interactive shells
+" and tools like Mutt and Vim pretty often. It feels natural to me to stack
+" issuing a :nohlsearch command to stop highlighting searches on top of this.
+"
+" This gets by far the most use in normal mode, but I’d like it to work in
+" insert and visual modes, too, where it’s occasionally useful, especially on
+" things like mobile phone terminal emulators that can be choppy and require
+" a lot of redrawing.
+"
+" For each of these, we end the mapping with a CTRL-L in normal mode, thereby
+" extending rather than replacing Vim’s normal behavior.
+"
nnoremap <C-L>
\ :<C-U>nohlsearch<CR><C-L>
+
+" The insert mode wrapper for normal CTRL-L uses i_CTRL-O to issue a single
+" normal mode command. We intentionally use ‘:normal’ rather than ‘:normal!’
+" so that the mapping works recursively. I tried using <C-O><C-L> with :imap
+" for this, but it didn’t work. Maybe i_CTRL-O doesn’t respect mappings.
+" I couldn’t find any documentation about it.
+"
inoremap <C-L> <C-O>:execute "normal \<C-L>"<CR>
+
+" We use :vnoremap here rather than :xnoremap and thereby make the mapping
+" apply to select mode as well, because CTRL-L doesn’t reflect a printable
+" character, and so we may as well make it work, even though I don’t actually
+" use select mode directly.
+"
vmap <C-L> <Esc><C-L>gv
+
+" By default, the very-useful normal mode command ‘&’ that repeats the
+" previous :substitute command doesn’t preserve the flags from that
+" substitution. I’d prefer it to do so, like the :&& command does, and it’s
+" easily remapped for both normal and visual mode, so let’s just do it.
+"
noremap &
\ :&&<CR>
ounmap &
sunmap &
+
+" I really like using the ‘!’ command in normal mode as an operator to filter
+" text through a shell command. It always bugged me a little that there
+" didn’t seem to be an analogue for a motion to filter text through an
+" internal command like :sort, so I wrote one.
+"
+" <https://sanctum.geek.nz/cgit/vim-colon-operator.git/about/>
+"
nmap g: <Plug>(ColonOperator)
+
+" I used Tim Pope’s unimpaired.vim plugin for ages, and I liked some of these
+" bracket pair mappings, so I’ve carried a few of the simpler ones over. All
+" of these can be prefixed with a count if needed, too. I use all of them
+" pretty regularly, even though cycling through lists to look for something
+" can be a bit wasteful.
+
+" Argument list
nnoremap [a
\ :previous<CR>
nnoremap ]a
\ :next<CR>
+" Buffers
nnoremap [b
\ :bprevious<CR>
nnoremap ]b
\ :bnext<CR>
+" Quickfix list
nnoremap [c
\ :cprevious<CR>
nnoremap ]c
\ :cnext<CR>
+" Location list
nnoremap [l
\ :lprevious<CR>
nnoremap ]l
\ :lnext<CR>
+
+" Here’s another mapping I particularly liked from unimpaired.vim; insert
+" blank lines from normal mode, using a custom plugin of mine called
+" put_blank_lines.vim. These use operator functions so that they’re
+" repeatable without repeat.vim. They accept count prefixes, too.
+"
+" <https://sanctum.geek.nz/cgit/vim-put-blank-lines.git/about/>
+"
nmap [<Space> <Plug>(PutBlankLinesAbove)
nmap ]<Space> <Plug>(PutBlankLinesBelow)
+
+" We’re on to the leader maps, now. It’s difficult to know in what order to
+" describe and specify these. I used to have them in alphabetical order, but
+" it seems much more useful to group them by the type of action they take.
+"
+" First of all, let’s set the leader keys; backslash happens to be the
+" default, but I like to make my choice explicit here. As of 2019, I’m still
+" not certain that comma is the best choice for my local leader. I use it all
+" the time for this purpose, and it works well, but I don’t much like that it
+" shadows a useful function in the fFtT;, group, and I sometimes wonder if
+" I would use the key for its original function more, had I not shadowed it.
+"
let mapleader = '\'
let maplocalleader = ','
+
+" If the local leader is a comma, map double-tap comma to its original
+" function in the relevant modes so that I can still use it quickly without
+" relying on mapping 'timeout'.
+"
if maplocalleader ==# ','
noremap ,, ,
sunmap ,,
endif
+
+" Let’s start with some simple ones; these ones all just toggle a boolean
+" option, and print its new value. They’re dirt simple to specify, and don’t
+" require any plugins.
+"
+" These are sometimes applicable in visual mode, and sometimes not. We’ll
+" start with the ones that only make sense as normal mode maps. Annoyingly,
+" a visual mode mapping for 'cursorline' toggling doesn’t work at all;
+" 'cursorline' is always off when in any visual mode, including block mode,
+" where it actually might have been really handy.
+
+"" Leader,TAB toggles automatic indentation based on the previous line
nnoremap <Leader><Tab>
\ :<C-U>set autoindent! autoindent?<CR>
+"" Leader,c toggles highlighted cursor row; doesn’t work in visual mode
nnoremap <Leader>c
\ :<C-U>set cursorline! cursorline?<CR>
+"" Leader,h toggles highlighting search results
nnoremap <Leader>h
\ :<C-U>set hlsearch! hlsearch?<CR>
+"" Leader,i toggles showing matches as I enter my pattern
nnoremap <Leader>i
\ :<C-U>set incsearch! incsearch?<CR>
+"" Leader,s toggles spell checking
nnoremap <Leader>s
\ :<C-U>set spell! spell?<CR>
+
+" The next group of option-toggling maps are much the same as the previous
+" group, except they also include analogous maps for visual mode, defined as
+" recursive maps into normal mode that conclude with re-selecting the text.
+
+"" Leader,C toggles highlighted cursor column; works in visual mode
noremap <Leader>C
\ :<C-U>set cursorcolumn! cursorcolumn?<CR>
ounmap <Leader>C
sunmap <Leader>C
+"" Leader,l toggles showing tab, end-of-line, and trailing white space
noremap <Leader>l
\ :<C-U>set list! list?<CR>
ounmap <Leader>l
sunmap <Leader>l
+"" Leader,n toggles line number display
noremap <Leader>n
\ :<C-U>set number! number?<CR>
ounmap <Leader>n
sunmap <Leader>n
+"" Leader,N toggles position display in bottom right
noremap <Leader>N
\ :<C-U>set ruler! ruler?<CR>
ounmap <Leader>N
sunmap <Leader>N
+"" Leader,w toggles soft wrapping
noremap <Leader>w
\ :<C-U>set wrap! wrap?<CR>
ounmap <Leader>w
sunmap <Leader>w
+
+" This next one just shows option state of the 'formatoptions' affecting how
+" text is automatically formatted; it doesn’t change its value.
+
+"" Leader,f shows the current 'formatoptions' at a glance
nnoremap <Leader>f
\ :<C-U>set formatoptions?<CR>
+
+" I often have to switch between US English and NZ English. The latter is
+" almost exactly the same as UK English in most locales, although we use
+" dollars rather than pounds. This is mostly so I remember things like
+" excluding or including the ‘u’ in words like 'favourite', depending on the
+" target audience. I generally use US English for international audiences.
+"
nnoremap <Leader>u
\ :<C-U>set spelllang=en_us<CR>
nnoremap <Leader>z
\ :<C-U>set spelllang=en_nz<CR>
+
+" The next mapping is also for toggling an option, but it’s more complicated;
+" it uses a simple plugin of mine called copy_linebreak.vim to manage several
+" options at once, related to the 'wrap' option that soft-wraps text.
+"
+" It’s designed for usage in terminal emulators and multiplexers to
+" temporarily make the buffer text suitable for copying in such a way that the
+" wrapping and any associated soft formatting won’t pervert the text,
+" including 'breakindent', 'linebreak', and 'showbreak' artifacts.
+"
+" This is really handy for quick selections of small regions of text. For
+" larger blocks of text or for manipulating the text as it leaves the buffer,
+" it makes more sense to use :! commands.
+"
+" <https://sanctum.geek.nz/cgit/vim-copy-linebreak.git/about/>
+"
+
+"" Leader,b toggles settings friendly to copying and pasting
nmap <Leader>b <Plug>(CopyLinebreakToggle)
+
+" The above mappings show that mappings for toggling boolean options are
+" simple, but there isn’t a way to toggle single flags within option strings
+" with just the :set command, so I wrote a plugin called toggle_flags.vim to
+" provide :ToggleFlag and :ToggleFlagLocal commands. The first argument is
+" the name of an option, and the second is the flag within it that should be
+" toggled on or off.
+
+"" Leader,a toggles 'formatoptions' ‘a’ auto-flowing flag
nnoremap <Leader>a
\ :<C-U>ToggleFlagLocal formatoptions a<CR>
+"" Leader,L toggles 'colorcolumn' showing the first column beyond 'textwidth'
noremap <Leader>L
\ :<C-U>ToggleFlagLocal colorcolumn +1<CR>
ounmap <Leader>L
sunmap <Leader>L
+
+" This mapping uses my paste_insert.vim plugin to queue up automatic commands
+" for the next insert operation. It’s still pretty new. It replaces my old
+" paste_open.vim plugin which did this only for opening new lines, and which
+" kept confusing me. I’m hoping this will be better.
+
+"" Leader,p prepares the next insert for paste mode
nmap <Leader>p <Plug>PasteInsert
+
+" These mappings are for managing filetypes. The first one uses the
+" :ReloadFileType command that was defined much earlier in this file for
+" application in the vimrc reload command.
+
+"" Leader,F reloads filetype settings
nnoremap <Leader>F
\ :<C-U>ReloadFileType<CR>
+"" Leader,t shows current filetype
nnoremap <Leader>t
\ :<C-U>set filetype?<CR>
+"" Leader,T clears filetype
nnoremap <Leader>T
\ :<C-U>set filetype=<CR>
+
+" These mappings use my put_date.vim plugin for date insertion into the
+" buffer.
+
+"" Leader,d inserts the local date (RFC 2822)
nnoremap <Leader>d
\ :PutDate<CR>
+"" Leader,D inserts the UTC date (RFC 2822)
nnoremap <Leader>D
\ :PutDate!<CR>
+
+" This group contains mappings that are to do with file and path management
+" relative to the current buffer. The Leader,P mapping that creates
+" directory hierarchies uses the :CreatePath command created earlier.
+
+"" Leader,g shows the current file’s fully expanded path
nnoremap <Leader>g
\ :<C-U>echo expand('%:p')<CR>
+"" Leader,G changes directory to the current file’s location
nnoremap <Leader>G
\ :<C-U>cd %:h<Bar>pwd<CR>
+"" Leader,P creates the path to the current file if it doesn’t exist
nnoremap <Leader>P
- \ :<C-U>Establish %:h<CR>
+ \ :<C-U>CreatePath %:h<CR>
+
+" This group contains mappings that show information about Vim’s internals:
+" marks, registers, variables, and the like.
+
+"" Leader,H shows command history
nnoremap <Leader>H
\ :<C-U>history :<CR>
+"" Leader,k shows my marks
nnoremap <Leader>k
\ :<C-U>marks<CR>
+"" Leader,K shows functions
nnoremap <Leader>K
\ :<C-U>function<CR>
+"" Leader,m shows normal maps
nnoremap <Leader>m
\ :<C-U>nmap<CR>
+"" Leader,M shows buffer-local normal maps
nnoremap <Leader>M
\ :<C-U>nmap <buffer><CR>
+"" Leader,S shows loaded scripts
nnoremap <Leader>S
\ :<C-U>scriptnames<CR>
+"" Leader,U shows user commands
nnoremap <Leader>U
\ :<C-U>command<CR>
+"" Leader,v shows all global and internal variables
nnoremap <Leader>v
\ :<C-U>let g: v:<CR>
+"" Leader,V shows all buffer, tab, and window local variables
nnoremap <Leader>V
\ :<C-U>let b: t: w:<CR>
+"" Leader,y shows all registers
nnoremap <Leader>y
\ :<C-U>registers<CR>
+
+" This group contains mappings concerned with buffer navigation and
+" management. I use the "jetpack" buffer jumper one a lot. I got it from one
+" of bairui’s "Vim and Vigor" comics:
+"
+" <http://of-vim-and-vigor.blogspot.com/p/vim-vigor-comic.html>
+
+"" Leader,DEL deletes the current buffer
nnoremap <Leader><Delete>
\ :bdelete<CR>
+"" Leader,INS edits a new buffer
nnoremap <Leader><Insert>
\ :<C-U>enew<CR>
+"" Leader,e forces a buffer to be editable, even a :help one
nnoremap <Leader>e
\ :<C-U>set modifiable noreadonly<CR>
+"" Leader,E locks a buffer, reversible with <Leader>e
nnoremap <Leader>E
\ :<C-U>set nomodifiable readonly<CR>
+"" Leader,j jumps to buffers ("jetpack")
nnoremap <Leader>j
\ :<C-U>buffers<CR>:buffer<Space>
+
+" Leader,o hacks up the list of old files from viminfo just long enough to
+" ensure that :browse :oldfiles fits in a screen, avoiding an Enter or ‘q’
+" keystroke before entering the number. This one is handy followed by
+" <Leader>,\ to jump back to the last remembered position in that file, since
+" by definition viminfo remembers that mark, too.
+"
nmap <Leader>o <Plug>(SelectOldFiles)
+
+" This group defines mappings for filtering and batch operations to clean up
+" buffer text. All of these mappings use commands from my custom plugins:
+"
+" :KeepPosition
+" <https://sanctum.geek.nz/cgit/vim-keep-position.git/about/>
+" :SqueezeRepeatBlanks
+" <https://sanctum.geek.nz/cgit/vim-squeeze-repeat-blanks.git/about/>
+" :StripTrailingWhitespace
+" <https://sanctum.geek.nz/cgit/vim-strip-trailing-whitespace.git/about/>
+"
+
+"" Leader,x strips trailing whitespace
noremap <Leader>x
\ :StripTrailingWhitespace<CR>
ounmap <Leader>x
sunmap <Leader>x
+
+"" Leader,X squeezes repeated blank lines
noremap <Leader>X
\ :SqueezeRepeatBlanks<CR>
ounmap <Leader>X
sunmap <Leader>X
+"" Leader,= runs the whole buffer through =, preserving position
nnoremap <Leader>=
\ :<C-U>KeepPosition execute 'normal! 1G=G'<CR>
+"" Leader,+ runs the whole buffer through gq, preserving position
nnoremap <Leader>+
\ :<C-U>KeepPosition execute 'normal! 1GgqG'<CR>
+
+" This group defines a few :onoremap commands to make my own text objects.
+" I should probably make some more of these, as they’ve proven to be
+" terrifically handy.
+
+"" Leader,_ uses last changed or yanked text as an object
onoremap <Leader>_
\ :<C-U>execute 'normal! `[v`]'<CR>
+"" Leader,% or Leader,5 uses entire buffer as an object
onoremap <Leader>%
\ :<C-U>execute 'normal! 1GVG'<CR>
omap <Leader>5 <Leader>%
+
+" This group defines some useful motions, including navigating by indent
+" block using a custom plugin:
+"
+" <https://sanctum.geek.nz/cgit/vim-vertical-region.git/about/>
+"
+
+"" Leader,{ and Leader,} move to top and bottom of indent region
map <Leader>{ <Plug>(VerticalRegionUp)
sunmap <Leader>{
map <Leader>} <Plug>(VerticalRegionDown)
sunmap <Leader>}
+"" Leader,\ jumps to the last edit position mark: think "Now, where was I?"
noremap <Leader>\ `"
sunmap <Leader>\
+
+" This group does both: useful motions on defined text objects.
+
+"" Leader,< and Leader,> adjust indent of last edit; good for pasting
nnoremap <Leader><lt>
\ :<C-U>'[,']<lt><CR>
nnoremap <Leader>>
\ :<C-U>'[,']><CR>
+
+" This group is for directory tree or help search convenience mappings.
+
+"" Leader,/ types :vimgrep for me ready to enter a search pattern
nnoremap <Leader>/
\ :<C-U>vimgrep /\c/j **<S-Left><S-Left><Right>
+"" Leader,? types :lhelpgrep for me ready to enter a search pattern
nnoremap <Leader>?
\ :<C-U>lhelpgrep \c<S-Left>
+
+" This group contains miscellaneous mappings for which I couldn’t find any
+" other place. The plugin mappings probably require their own documentation
+" comment block, but my hands are getting tired from all this typing.
+"
+" * <https://sanctum.geek.nz/cgit/vim-alternate-filetype.git/about/>
+" * <https://sanctum.geek.nz/cgit/vim-regex-escape.git/about/>
+" * <https://sanctum.geek.nz/cgit/vim-replace-operator.git/about/>
+" * <https://sanctum.geek.nz/cgit/vim-scratch-buffer.git/about/>
+"
+
+"" Leader,. runs the configured make program into the location list
nnoremap <Leader>.
\ :<C-U>lmake!<CR>
+"" Leader,q formats the current paragraph
nnoremap <Leader>q gqap
+"" Leader,r acts as a replacement operator
map <Leader>r <Plug>(ReplaceOperator)
-sunmap <Leader>r
ounmap <Leader>r
+sunmap <Leader>r
+"" Leader,!/1 repeats the last command, adding a bang
nnoremap <Leader>!
\ :<Up><Home><S-Right>!<CR>
nmap <Leader>1 <Leader>!
+"" Leader,#/3 gives me my fortune
nmap <Leader># <Plug>(AlternateFileType)
nmap <Leader>3 <Leader>#
+"" Leader,$/4 gives me my fortune
nmap <Leader>$ <Plug>(Fortune)
nmap <Leader>4 <Leader>$
-nmap <Leader>& <Plug>(RegexEscape)
-nmap <Leader>7 <Leader>&
-xmap <Leader>& <Plug>(RegexEscape)
-xmap <Leader>7 <Leader>&
+"" Leader,&/7 escapes regex metacharacters
+map <Leader>& <Plug>(RegexEscape)
+ounmap <Leader>&
+sunmap <Leader>&
+map <Leader>7 <Leader>&
+ounmap <Leader>7
+sunmap <Leader>7
+"" Leader,*/8 is sticky-star: search, highlight, but don’t move
nnoremap <silent> <Leader>* *N
nmap <Leader>8 <Leader>*
+"" Leader,` opens a scratch buffer, horizontally split
nnoremap <silent> <Leader>`
\ :<C-U>ScratchBuffer<CR>
+"" Leader,~ opens a scratch buffer, vertically split
nnoremap <silent> <Leader>~
\ :<C-U>vertical ScratchBuffer<CR>
+
+" And last, but definitely not least, I’m required by Vim fanatic law to
+" include a mapping that reloads my whole configuration. This uses the
+" command wrapper defined much earlier in the file, so that filetypes also get
+" reloaded afterwards, meaning I don’t need to follow <Leader>R with
+" a <Leader>F to fix up broken global settings.
+"
nnoremap <Leader>R
\ :<C-U>ReloadVimrc<CR>
+
+" I’ll close this file with a few abbreviations. Perhaps of everything in
+" here, I’m least confident that these should be in here, but they’ve proven
+" pretty useful. First, some 'deliberate' abbreviations for stuff I type
+" a lot:
+"
inoreabbrev tr@ tom@sanctum.geek.nz
inoreabbrev tr/ <https://sanctum.geek.nz/>
+
+" And then, just automatically fix some things I almsot always spell or type
+" wrnog.
+"
inoreabbrev almsot almost
inoreabbrev wrnog wrong
inoreabbrev Fielding Feilding