Nvim :help pages, generated from source using the tree-sitter-vimdoc parser.

Spell checking
This command switches on spell checking:
:setlocal spell spelllang=en_us
This switches on the 'spell' option and specifies to check for US English.
The words that are not recognized are highlighted with one of these: SpellBad word not recognized hl-SpellBad SpellCap word not capitalised hl-SpellCap SpellRare rare word hl-SpellRare SpellLocal wrong spelling for selected region hl-SpellLocal
Vim only checks words for spelling, there is no grammar check.
If the 'mousemodel' option is set to "popup" and the cursor is on a badly spelled word or it is "popup_setpos" and the mouse pointer is on a badly spelled word, then the popup menu will contain a submenu to replace the bad word. Note: this slows down the appearance of the popup menu.
To search for the next misspelled word:
]s ]s Move to next misspelled word after the cursor. A count before the command can be used to repeat. 'wrapscan' applies.
[s [s Like "]s" but search backwards, find the misspelled word before the cursor. Doesn't recognize words split over two lines, thus may stop at words that are not highlighted as bad. Does not stop at word with missing capital at the start of a line.
]S ]S Like "]s" but only stop at bad words, not at rare words or words for another region.
[S [S Like "]S" but search backwards.
]r ]r Move to next "rare" word after the cursor. A count before the command can be used to repeat. 'wrapscan' applies.
[r [r Like "]r" but search backwards, find the "rare" word before the cursor. Doesn't recognize words split over two lines, thus may stop at words that are not highlighted as rare.
To add words to your own word list:
zg zg Add word under the cursor as a good word to the first name in 'spellfile'. A count may precede the command to indicate the entry in 'spellfile' to be used. A count of two uses the second entry.
In Visual mode the selected characters are added as a word (including white space!). When the cursor is on text that is marked as badly spelled then the marked text is used. Otherwise the word under the cursor, separated by non-word characters, is used.
If the word is explicitly marked as bad word in another spell file the result is unpredictable.
zG zG Like "zg" but add the word to the internal word list internal-wordlist.
zw zw Like "zg" but mark the word as a wrong (bad) word. If the word already appears in 'spellfile' it is turned into a comment line. See spellfile-cleanup for getting rid of those.
zW zW Like "zw" but add the word to the internal word list internal-wordlist.
zuw zug zuw zug Undo zw and zg, remove the word from the entry in 'spellfile'. Count used as with zg.
zuW zuG zuW zuG Undo zW and zG, remove the word from the internal word list. Count used as with zg.
:spe :spellgood E1280 :[count]spe[llgood] {word} Add {word} as a good word to 'spellfile', like with zg. Without count the first name is used, with a count of two the second entry, etc.
:spe[llgood]! {word} Add {word} as a good word to the internal word list, like with zG.
:spellw :spellwrong :[count]spellw[rong] {word} Add {word} as a wrong (bad) word to 'spellfile', as with zw. Without count the first name is used, with a count of two the second entry, etc.
:spellw[rong]! {word} Add {word} as a wrong (bad) word to the internal word list, like with zW.
:spellra :spellrare :[count]spellra[re] {word} Add {word} as a rare word to 'spellfile', similar to zw. Without count the first name is used, with a count of two the second entry, etc.
There are no normal mode commands to mark words as rare as this is a fairly uncommon command and all intuitive commands for this are already taken. If you want you can add mappings with e.g.:
nnoremap z?  :exe ':spellrare  ' .. expand('<cWORD>')<CR>
nnoremap z/  :exe ':spellrare! ' .. expand('<cWORD>')<CR>
:spellundo, zuw, or zuW can be used to undo this.
:spellra[re]! {word} Add {word} as a rare word to the internal word list, similar to zW.
:[count]spellu[ndo] {word} :spellu :spellundo Like zuw. [count] used as with :spellgood.
:spellu[ndo]! {word} Like zuW. [count] used as with :spellgood.
After adding a word to 'spellfile' with the above commands its associated ".spl" file will automatically be updated and reloaded. If you change 'spellfile' manually you need to use the :mkspell command. This sequence of commands mostly works well:
:edit <file in 'spellfile'>
(make changes to the spell file)
:mkspell! %
More details about the 'spellfile' format below spell-wordlist-format.
internal-wordlist The internal word list is used for all buffers where 'spell' is set. It is not stored, it is lost when you exit Vim. It is also cleared when 'encoding' is set.
Finding suggestions for bad words: z= z= For the word under/after the cursor suggest correctly spelled words. This also works to find alternatives for a word that is not highlighted as a bad word, e.g., when the word after it is bad. In Visual mode the highlighted text is taken as the word to be replaced. The results are sorted on similarity to the word being replaced. This may take a long time. Hit CTRL-C when you get bored.
If the command is used without a count the alternatives are listed and you can enter the number of your choice or press <Enter> if you don't want to replace. You can also use the mouse to click on your choice (only works if the mouse can be used in Normal mode and when there are no line wraps). Click on the first line (the header) to cancel.
The suggestions listed normally replace a highlighted bad word. Sometimes they include other text, in that case the replaced text is also listed after a "<".
If a count is used that suggestion is used, without prompting. For example, "1z=" always takes the first suggestion.
If 'verbose' is non-zero a score will be displayed with the suggestions to indicate the likeliness to the badly spelled word (the higher the score the more different). When a word was replaced the redo command "." will repeat the word replacement. This works like "ciw", the good word and <Esc>. This does NOT work for Thai and other languages without spaces between words.
:spellr :spellrepall E752 E753 :spellr[epall] Repeat the replacement done by z= for all matches with the replaced word in the current window.
In Insert mode, when the cursor is after a badly spelled word, you can use CTRL-X s to find suggestions. This works like Insert mode completion. Use CTRL-N to use the next suggestion, CTRL-P to go back. i_CTRL-X_s
The 'spellsuggest' option influences how the list of suggestions is generated and sorted. See 'spellsuggest'.
The 'spellcapcheck' option is used to check the first word of a sentence starts with a capital. This doesn't work for the first word in the file. When there is a line break right after a sentence the highlighting of the next line may be postponed. Use CTRL-L when needed. Also see set-spc-auto for how it can be set automatically when 'spelllang' is set.
The 'spelloptions' option has a few more flags that influence the way spell checking works. For example, "camel" splits CamelCased words so that each part of the word is spell-checked separately.
Vim counts the number of times a good word is encountered. This is used to sort the suggestions: words that have been seen before get a small bonus, words that have been seen often get a bigger bonus. The COMMON item in the affix file can be used to define common words, so that this mechanism also works in a new or short file spell-COMMON.
Vim does on-the-fly spell checking. To make this work fast the word list is loaded in memory. Thus this uses a lot of memory (1 Mbyte or more). There might also be a noticeable delay when the word list is loaded, which happens when 'spell' is set and when 'spelllang' is set while 'spell' was already set. To minimize the delay each word list is only loaded once, it is not deleted when 'spelllang' is made empty or 'spell' is reset. When 'encoding' is set all the word lists are reloaded, thus you may notice a delay then too.
A word may be spelled differently in various regions. For example, English comes in (at least) these variants:
en all regions en_au Australia en_ca Canada en_gb Great Britain en_nz New Zealand en_us USA
Words that are not used in one region but are used in another region are highlighted with SpellLocal hl-SpellLocal.
Always use lowercase letters for the language and region names.
When adding a word with zg or another command it's always added for all regions. You can change that by manually editing the 'spellfile'. See spell-wordlist-format. Note that the regions as specified in the files in 'spellfile' are only used when all entries in 'spelllang' specify the same region (not counting files specified by their .spl name).
spell-german Specific exception: For German these special regions are used: de all German words accepted de_de old and new spelling de_19 old spelling de_20 new spelling de_at Austria de_ch Switzerland
spell-russian Specific exception: For Russian these special regions are used: ru all Russian words accepted ru_ru "IE" letter spelling ru_yo "YO" letter spelling
spell-yiddish Yiddish requires using "utf-8" encoding, because of the special characters used. If you are using latin1 Vim will use transliterated (romanized) Yiddish instead. If you want to use transliterated Yiddish with utf-8 use "yi-tr". In a table: 'encoding' 'spelllang' utf-8 yi Yiddish latin1 yi transliterated Yiddish utf-8 yi-tr transliterated Yiddish
spell-cjk Chinese, Japanese and other East Asian characters are normally marked as errors, because spell checking of these characters is not supported. If 'spelllang' includes "cjk", these characters are not marked as errors. This is useful when editing text with spell checking while some Asian words are present.
Vim searches for spell files in the "spell" subdirectory of the directories in 'runtimepath'. The name is: LL.EEE.spl, where: LL the language name EEE the value of 'encoding'
The value for "LL" comes from 'spelllang', but excludes the region name. Examples: en_us en en-rare en-rare medical_ca medical
Only the first file is loaded, the one that is first in 'runtimepath'. If this succeeds then additionally files with the name LL.EEE.add.spl are loaded. All the ones that are found are used.
If no spell file is found the SpellFileMissing autocommand event is triggered. This may trigger the spellfile.vim plugin to offer you downloading the spell file.
Additionally, the files related to the names in 'spellfile' are loaded. These are the files that zg and zw add good and wrong words to.
Vim uses "latin1" when 'encoding' is "iso-8859-15". The euro sign doesn't matter for spelling.
When no spell file for 'encoding' is found "ascii" is tried. This only works for languages where nearly all words are ASCII, such as English. It helps when 'encoding' is not "latin1", such as iso-8859-2, and English text is being edited. For the ".add" files the same name as the found main spell file is used.
For example, with these values: 'runtimepath' is "~/.config/nvim,/usr/share/nvim/runtime/,~/.config/nvim/after" 'encoding' is "iso-8859-2" 'spelllang' is "pl"
Vim will look for: 1. ~/.config/nvim/spell/pl.iso-8859-2.spl 2. /usr/share/nvim/runtime/spell/pl.iso-8859-2.spl 3. ~/.config/nvim/spell/pl.iso-8859-2.add.spl 4. /usr/share/nvim/runtime/spell/pl.iso-8859-2.add.spl 5. ~/.config/nvim/after/spell/pl.iso-8859-2.add.spl
This assumes 1. is not found and 2. is found.
If 'encoding' is "latin1" Vim will look for: 1. ~/.config/nvim/spell/pl.latin1.spl 2. /usr/share/nvim/runtime/spell/pl.latin1.spl 3. ~/.config/nvim/after/spell/pl.latin1.spl 4. ~/.config/nvim/spell/pl.ascii.spl 5. /usr/share/nvim/runtime/spell/pl.ascii.spl 6. ~/.config/nvim/after/spell/pl.ascii.spl
This assumes none of them are found (Polish doesn't make sense when leaving out the non-ASCII characters).
A spell file might not be available in the current 'encoding'. See spell-mkspell about how to create a spell file. Converting a spell file with "iconv" will NOT work!
spell-sug-file E781 If there is a file with exactly the same name as the ".spl" file but ending in ".sug", that file will be used for giving better suggestions. It isn't loaded before suggestions are made to reduce memory use.
E758 E759 E778 E779 E780 E782 When loading a spell file Vim checks that it is properly formatted. If you get an error the file may be truncated, modified or intended for another Vim version.
The zw command turns existing entries in 'spellfile' into comment lines. This avoids having to write a new file every time, but results in the file only getting longer, never shorter. To clean up the comment lines in all ".add" spell files do this:
:runtime spell/cleanadd.vim
This deletes all comment lines, except the ones that start with "##". Use "##" lines to add comments that you want to keep.
You can invoke this script as often as you like. A variable is provided to skip updating files that have been changed recently. Set it to the number of seconds that has passed since a file was changed before it will be cleaned. For example, to clean only files that were not changed in the last hour:
let g:spell_clean_limit = 60 * 60
The default is one second.
Vim uses a fixed method to recognize a word. This is independent of 'iskeyword', so that it also works in help files and for languages that include characters like '-' in 'iskeyword'. The word characters do depend on 'encoding'.
The table with word characters is stored in the main .spl file. Therefore it matters what the current locale is when generating it! A .add.spl file does not contain a word table though.
For a word that starts with a digit the digit is ignored, unless the word as a whole is recognized. Thus if "3D" is a word and "D" is not then "3D" is recognized as a word, but if "3D" is not a word then only the "D" is marked as bad. Hex numbers in the form 0x12ab and 0X12AB are recognized.
It is possible to spell-check words that include a space. This is used to recognize words that are invalid when used by themselves, e.g. for "et al.". It can also be used to recognize "the the" and highlight it.
The number of spaces is irrelevant. In most cases a line break may also appear. However, this makes it difficult to find out where to start checking for spelling mistakes. When you make a change to one line and only that line is redrawn Vim won't look in the previous line, thus when "et" is at the end of the previous line "al." will be flagged as an error. And when you type "the<CR>the" the highlighting doesn't appear until the first line is redrawn. Use CTRL-L to redraw right away. "[s" will also stop at a word combination with a line break.
When encountering a line break Vim skips characters such as "*", '>' and '"', so that comments in C, shell and Vim code can be spell checked.
Files that use syntax highlighting can specify where spell checking should be done:
1. everywhere default 2. in specific items use "contains=@Spell" 3. everywhere but specific items use "contains=@NoSpell"
For the second method adding the @NoSpell cluster will disable spell checking again. This can be used, for example, to add @Spell to the comments of a program, and add @NoSpell for items that shouldn't be checked. Also see :syn-spell for text that is not in a syntax item.
If you want to write a Vim script that does something with spelling, you may find these functions useful:
spellbadword() find badly spelled word at the cursor spellsuggest() get list of spelling suggestions soundfold() get the sound-a-like version of a word
SETTING 'spellcapcheck' AUTOMATICALLY set-spc-auto
After the 'spelllang' option has been set successfully, Vim will source the files "spell/LANG.vim" and "spell/LANG.lua" in 'runtimepath'. "LANG" is the value of 'spelllang' up to the first comma, dot or underscore. This can be used to set options specifically for the language, especially 'spellcapcheck'.
The distribution includes a few of these files. Use this command to see what they do:
:next $VIMRUNTIME/spell/*.vim
Note that the default scripts don't set 'spellcapcheck' if it was changed from the default value. This assumes the user prefers another value then.
The 'spellsuggest' option can be used to select "double" scoring. This mechanism is based on the principle that there are two kinds of spelling mistakes:
1. You know how to spell the word, but mistype something. This results in a small editing distance (character swapped/omitted/inserted) and possibly a word that sounds completely different.
2. You don't know how to spell the word and type something that sounds right. The edit distance can be big but the word is similar after sound-folding.
Since scores for these two mistakes will be very different we use a list for each and mix them.
The sound-folding is slow and people that know the language won't make the second kind of mistakes. Therefore 'spellsuggest' can be set to select the preferred method for scoring the suggestions.
Vim uses a binary file format for spelling. This greatly speeds up loading the word list and keeps it small. .aff .dic Myspell You can create a Vim spell file from the .aff and .dic files that Myspell uses. Myspell is used by OpenOffice.org and Mozilla. The OpenOffice .oxt files are zip files which contain the .aff and .dic files. You should be able to find them here: https://extensions.services.openoffice.org/dictionary The older, OpenOffice 2 files may be used if this doesn't work: http://wiki.services.openoffice.org/wiki/Dictionaries You can also use a plain word list. The results are the same, the choice depends on what word lists you can find.
If you install Aap (from www.a-a-p.org) you can use the recipes in the runtime/spell/??/ directories. Aap will take care of downloading the files, apply patches needed for Vim and build the .spl file.
Make sure your current locale is set properly, otherwise Vim doesn't know what characters are upper/lower case letters. If the locale isn't available (e.g., when using an MS-Windows codepage on Unix) add tables to the .aff file spell-affix-chars. If the .aff file doesn't define a table then the word table of the currently active spelling is used. If spelling is not active then Vim will try to guess.
:mksp :mkspell :mksp[ell][!] [-ascii] {outname} {inname} ... Generate a Vim spell file from word lists. Example:
:mkspell /tmp/nl nl_NL.words
E751 When {outname} ends in ".spl" it is used as the output file name. Otherwise it should be a language name, such as "en", without the region name. The file written will be "{outname}.{encoding}.spl", where {encoding} is the value of the 'encoding' option.
When the output file already exists [!] must be used to overwrite it.
When the [-ascii] argument is present, words with non-ascii characters are skipped. The resulting file ends in "ascii.spl".
The input can be the Myspell format files {inname}.aff and {inname}.dic. If {inname}.aff does not exist then {inname} is used as the file name of a plain word list.
Multiple {inname} arguments can be given to combine regions into one Vim spell file. Example:
:mkspell ~/.config/nvim/spell/en /tmp/en_US /tmp/en_CA /tmp/en_AU
This combines the English word lists for US, CA and AU into one en.spl file. Up to eight regions can be combined. E754 E755 The REP and SAL items of the first .aff file where they appear are used. spell-REP spell-SAL E845 This command uses a lot of memory, required to find the optimal word tree (Polish, Italian and Hungarian require several hundred Mbyte). The final result will be much smaller, because compression is used. To avoid running out of memory compression will be done now and then. This can be tuned with the 'mkspellmem' option.
After the spell file was written and it was being used in a buffer it will be reloaded automatically.
:mksp[ell] [-ascii] {name}.{enc}.add Like ":mkspell" above, using {name}.{enc}.add as the input file and producing an output file in the same directory that has ".spl" appended.
:mksp[ell] [-ascii] {name} Like ":mkspell" above, using {name} as the input file and producing an output file in the same directory that has ".{enc}.spl" appended.
Vim will report the number of duplicate words. This might be a mistake in the list of words. But sometimes it is used to have different prefixes and suffixes for the same basic word to avoid them combining (e.g. Czech uses this). If you want Vim to report all duplicate words set the 'verbose' option.
Since you might want to change a Myspell word list for use with Vim the following procedure is recommended:
1. Obtain the xx_YY.aff and xx_YY.dic files from Myspell. 2. Make a copy of these files to xx_YY.orig.aff and xx_YY.orig.dic. 3. Change the xx_YY.aff and xx_YY.dic files to remove bad words, add missing words, define word characters with FOL/LOW/UPP, etc. The distributed "*.diff" files can be used. 4. Start Vim with the right locale and use :mkspell to generate the Vim spell file. 5. Try out the spell file with ":set spell spelllang=xx" if you wrote it in a spell directory in 'runtimepath', or ":set spelllang=xx.enc.spl" if you wrote it somewhere else.
When the Myspell files are updated you can merge the differences: 1. Obtain the new Myspell files as xx_YY.new.aff and xx_UU.new.dic. 2. Use diff-mode to see what changed:
nvim -d xx_YY.orig.dic xx_YY.new.dic
3. Take over the changes you like in xx_YY.dic. You may also need to change xx_YY.aff. 4. Rename xx_YY.new.dic to xx_YY.orig.dic and xx_YY.new.aff to xx_YY.orig.aff.
Spell checking is a relatively new feature in Vim, thus it's possible that the .spl file format will be changed to support more languages. Vim will check the validity of the spell file and report anything wrong.
This spell file is older than your Vim. You need to update the .spl file.
This means the spell file was made for a later version of Vim. You need to update Vim.
This means the spell file was made for a later version of Vim and contains a section that is required for the spell file to work. In this case it's probably a good idea to upgrade your Vim.
If for some reason you want to check what words are supported by the currently used spelling files, use this command:
:spelldump :spelld :spelld[ump] Open a new window and fill it with all currently valid words. Compound words are not included. Note: For some languages the result may be enormous, causing Vim to run out of memory.
:spelld[ump]! Like ":spelldump" and include the word count. This is the number of times the word was found while updating the screen. Words that are in COMMON items get a starting count of 10.
The format of the word list is used spell-wordlist-format. You should be able to read it with ":mkspell" to generate one .spl file that includes all the words.
When all entries to 'spelllang' use the same regions or no regions at all then the region information is included in the dumped words. Otherwise only words for the current region are included and no "/regions" line is generated.
Comment lines with the name of the .spl file are used as a header above the words that were generated from that .spl file.
If the spell file for the language you are using is not available, you will get an error message. But if the "spellfile.vim" plugin is active it will offer you to download the spell file. Just follow the instructions, it will ask you where to write the file (there must be a writable directory in 'runtimepath' for this).
The plugin has a default place where to look for spell files, on the Vim ftp server. The protocol used is SSL (https://) for security. If you want to use another location or another protocol, set the g:spellfile_URL variable to the directory that holds the spell files. You can use http:// or ftp://, but you are taking a security risk then. The netrw plugin is used for getting the file, look there for the specific syntax of the URL. Example:
let g:spellfile_URL = 'https://ftp.nluug.nl/vim/runtime/spell'
You may need to escape special characters.
The plugin will only ask about downloading a language once. If you want to try again anyway restart Vim, or set g:spellfile_URL to another value (e.g., prepend a space).
To avoid using the "spellfile.vim" plugin do this in your vimrc file:
let loaded_spellfile_plugin = 1
Instead of using the plugin you can define a SpellFileMissing autocommand to handle the missing file yourself. You can use it like this:
:au SpellFileMissing * call Download_spell_file(expand('<amatch>'))
Thus the <amatch> item contains the name of the language. Another important value is 'encoding', since every encoding has its own spell file. With two exceptions:
For ISO-8859-15 (latin9) the name "latin1" is used (the encodings only differ in characters not used in dictionary words).
The name "ascii" may also be used for some languages where the words use only ASCII letters for most of the words.
The default "spellfile.vim" plugin uses this autocommand, if you define your autocommand afterwards you may want to use ":au! SpellFileMissing" to overrule it. If you define your autocommand before the plugin is loaded it will notice this and not do anything. E797 Note that the SpellFileMissing autocommand must not change or destroy the buffer the user was editing.
This is the format of the files that are used by the person who creates and maintains a word list.
Note that we avoid the word "dictionary" here. That is because the goal of spell checking differs from writing a dictionary (as in the book). For spelling we need a list of words that are OK, thus should not be highlighted. Person and company names will not appear in a dictionary, but do appear in a word list. And some old words are rarely used while they are common misspellings. These do appear in a dictionary but not in a word list.
There are two formats: A straight list of words and a list using affix compression. The files with affix compression are used by Myspell (Mozilla and OpenOffice.org). This requires two files, one with .aff and one with .dic extension.
The words must appear one per line. That is all that is required.
Additionally the following items are recognized:
Empty and blank lines are ignored.
Lines starting with a # are ignored (comment lines).
A line starting with "/encoding=", before any word, specifies the encoding of the file. After the second '=' comes an encoding name. This tells Vim to setup conversion from the specified encoding to 'encoding'. Thus you can use one word list for several target encodings.
A line starting with "/regions=" specifies the region names that are supported. Each region name must be two ASCII letters. The first one is region 1. Thus "/regions=usca" has region 1 "us" and region 2 "ca". In an addition word list the region names should be equal to the main word list!
Other lines starting with '/' are reserved for future use. The ones that are not recognized are ignored. You do get a warning message, so that you know something won't work.
A "/" may follow the word with the following items: = Case must match exactly. ? Rare word. ! Bad (wrong) word. 1 to 9 A region in which the word is valid. If no regions are specified the word is valid in all regions.
# This is an example word list comment /encoding=latin1 encoding of the file /regions=uscagb regions "us", "ca" and "gb" example word for all regions blah/12 word for regions "us" and "ca" vim/! bad word Campbell/?3 rare word in region 3 "gb" 's mornings/= keep-case word
Note that when "/=" is used the same word with all upper-case letters is not accepted. This is different from a word with mixed case that is automatically marked as keep-case, those words may appear in all upper-case letters.
There are two files: the basic word list and an affix file. The affix file specifies settings for the language and can contain affixes. The affixes are used to modify the basic words to get the full word list. This significantly reduces the number of words, especially for a language like Polish. This is called affix compression.
The basic word list and the affix file are combined with the ":mkspell" command and results in a binary spell file. All the preprocessing has been done, thus this file loads fast. The binary spell file format is described in the source code (src/spell.c). But only developers need to know about it.
The preprocessing also allows us to take the Myspell language files and modify them before the Vim word list is made. The tools for this can be found in the "src/spell" directory.
The format for the affix and word list files is based on what Myspell uses (the spell checker of Mozilla and OpenOffice.org). A description can be found here: Note that affixes are case sensitive, this isn't obvious from the description.
Vim supports quite a few extras. They are described below spell-affix-vim. Attempts have been made to keep this compatible with other spell checkers, so that the same files can often be used. One other project that offers more than Myspell is Hunspell ( https://hunspell.github.io ).
A short example, with line numbers:
The first line contains the number of words. Vim ignores it, but you do get an error message if it's not there. E760
What follows is one word per line. White space at the end of the line is ignored, all other white space matters. The encoding is specified in the affix file spell-SET.
Comment lines start with '#' or '/'. See the example lines 8 and 12. Note that putting a comment after a word is NOT allowed:
After the word there is an optional slash and flags. Most of these flags are letters that indicate the affixes that can be used with this word. These are specified with SFX and PFX lines in the .aff file, see spell-SFX and spell-PFX. Vim allows using other flag types with the FLAG item in the affix file spell-FLAG.
When the word only has lower-case letters it will also match with the word starting with an upper-case letter.
When the word includes an upper-case letter, this means the upper-case letter is required at this position. The same word with a lower-case letter at this position will not match. When some of the other letters are upper-case it will not match either.
The word with all upper-case characters will always be OK,
als als Als ALS ALs AlS aLs aLS Als Als ALS als ALs AlS aLs aLS ALS ALS als Als ALs AlS aLs aLS AlS AlS ALS als Als ALs aLs aLS
The KEEPCASE affix ID can be used to specifically match a word with identical case only, see below spell-KEEPCASE.
Note: in line 5 to 7 non-word characters are used. You can include any character in a word. When checking the text a word still only matches when it appears with a non-word character before and after it. For Myspell a word starting with a non-word character probably won't work.
In line 12 the word "TCP/IP" is defined. Since the slash has a special meaning the comma is used instead. This is defined with the SLASH item in the affix file, see spell-SLASH. Note that without this SLASH item the word will be "TCP,IP".
spell-affix-comment Comment lines in the .aff file start with a '#':
Items with a fixed number of arguments can be followed by a comment. But only if none of the arguments can contain white space. The comment must start with a "#" character. Example:
The affix file can be in any encoding that is supported by "iconv". However, in some cases the current locale should also be set properly at the time :mkspell is invoked. Adding FOL/LOW/UPP lines removes this requirement spell-FOL.
The encoding should be specified before anything where the encoding matters. The encoding applies both to the affix file and the dictionary file. It is done with a SET line:
The encoding can be different from the value of the 'encoding' option at the time ":mkspell" is used. Vim will then convert everything to 'encoding' and generate a spell file for 'encoding'. If some of the used characters to not fit in 'encoding' you will get an error message. spell-affix-mbyte When using a multibyte encoding it's possible to use more different affix flags. But Myspell doesn't support that, thus you may not want to use it anyway. For compatibility use an 8-bit encoding.
These entries in the affix file can be used to add information to the spell file. There are no restrictions on the format, but they should be in the right encoding.
spell-NAME spell-VERSION spell-HOME spell-AUTHOR spell-EMAIL spell-COPYRIGHT NAME Name of the language VERSION 1.0.1 with fixes HOME https://www.example.com AUTHOR John Doe EMAIL john AT Doe DOT net COPYRIGHT LGPL
These fields are put in the .spl file as-is. The :spellinfo command can be used to view the info.
:spellinfo :spelli :spelli[nfo] Display the information for the spell file(s) used for the current buffer.


spell-affix-chars When using an 8-bit encoding the affix file should define what characters are word characters. This is because the system where ":mkspell" is used may not support a locale with this encoding and isalpha() won't work. For example when using "cp1250" on Unix. E761 E762 spell-FOL spell-LOW spell-UPP Three lines in the affix file are needed. Simplistic example:
All three lines must have exactly the same number of characters.
The "FOL" line specifies the case-folded characters. These are used to compare words while ignoring case. For most encodings this is identical to the lower case line.
The "LOW" line specifies the characters in lower-case. Mostly it's equal to the "FOL" line.
The "UPP" line specifies the characters with upper-case. That is, a character is upper-case where it's different from the character at the same position in "FOL".
An exception is made for the German sharp s ß. The upper-case version is "SS". In the FOL/LOW/UPP lines it should be included, so that it's recognized as a word character, but use the ß character in all three.
ASCII characters should be omitted, Vim always handles these in the same way. When the encoding is UTF-8 no word characters need to be specified.
E763 Vim allows you to use spell checking for several languages in the same file. You can list them in the 'spelllang' option. As a consequence all spell files for the same encoding must use the same word characters, otherwise they can't be combined without errors.
If you get an E763 warning that the word tables differ you need to update your ".spl" spell files. If you downloaded the files, get the latest version of all spell files you use. If you are only using one, e.g., German, then also download the recent English spell files. Otherwise generate the .spl file again with :mkspell. If you still get errors check the FOL, LOW and UPP lines in the used .aff files.
The XX.ascii.spl spell file generated with the "-ascii" argument will not contain the table with characters, so that it can be combine with spell files for any encoding. The .add.spl files also do not contain the table.


spell-midword Some characters are only to be considered word characters if they are used in between two ordinary word characters. An example is the single quote: It is often used to put text in quotes, thus it can't be recognized as a word character, but when it appears in between word characters it must be part of the word. This is needed to detect a spelling error such as they'are. That should be they're, but since "they" and "are" are words themselves that would go unnoticed.
These characters are defined with MIDWORD in the .aff file. Example:
Flags are used to specify the affixes that can be used with a word and for other properties of the word. Normally single-character flags are used. This limits the number of possible flags, especially for 8-bit encodings. The FLAG item can be used if more affixes are to be used. Possible values:
FLAG long use two-character flags FLAG num use numbers, from 1 up to 65000 FLAG caplong use one-character flags without A-Z and two-character flags that start with A-Z
With "FLAG num" the numbers in a list of affixes need to be separated with a comma: "234,2143,1435". This method is inefficient, but useful if the file is generated with a program.
When using "caplong" the two-character flags all start with a capital: "Aa", "B1", "BB", etc. This is useful to use one-character flags for the most common items and two-character flags for uncommon items.
Note: When using utf-8 only characters up to 65000 may be used for flags.
Note: even when using "num" or "long" the number of flags available to compounding and prefixes is limited to about 250.
The usual PFX (prefix) and SFX (suffix) lines are supported (see the Myspell documentation or the Aspell manual: http://aspell.net/man-html/Affix-Compression.html).
The first line is a header and has four fields: SFX {flag} {combine} {count}
{flag} The name used for the suffix. Mostly it's a single letter, but other characters can be used, see spell-FLAG.
{combine} Can be 'Y' or 'N'. When 'Y' then the word plus suffix can also have a prefix. When 'N' then a prefix is not allowed.
{count} The number of lines following. If this is wrong you will get an error message.
For PFX the fields are exactly the same.
The basic format for the following lines is: SFX {flag} {strip} {add} {condition} {extra}
{flag} Must be the same as the {flag} used in the first line.
{strip} Characters removed from the basic word. There is no check if the characters are actually there, only the length is used (in bytes). This better match the {condition}, otherwise strange things may happen. If the {strip} length is equal to or longer than the basic word the suffix won't be used. When {strip} is 0 (zero) then nothing is stripped.
{add} Characters added to the basic word, after removing {strip}. Optionally there is a '/' followed by flags. The flags apply to the word plus affix. See spell-affix-flags
{condition} A simplistic pattern. Only when this matches with a basic word will the suffix be used for that word. This is normally for using one suffix letter with different {add} and {strip} fields for words with different endings. When {condition} is a . (dot) there is no condition. The pattern may contain:
Literal characters.
A set of characters in []. [abc] matches a, b and c. A dash is allowed for a range [a-c], but this is Vim-specific.
A set of characters that starts with a ^, meaning the complement of the specified characters. [^abc] matches any character but a, b and c.
{extra} Optional extra text: # comment Comment is ignored - Hunspell uses this, ignored
For PFX the fields are the same, but the {strip}, {add} and {condition} apply to the start of the word.
Note: Myspell ignores any extra text after the relevant info. Vim requires this text to start with a "#" so that mistakes don't go unnoticed. Example:
However, to avoid lots of errors in affix files written for Myspell, you can add the IGNOREEXTRA flag.
Apparently Myspell allows an affix name to appear more than once. Since this might also be a mistake, Vim checks for an extra "S". The affix files for Myspell that use this feature apparently have this flag. Example:
This is a feature that comes from Hunspell: The affix may specify flags. This works similar to flags specified on a basic word. The flags apply to the basic word plus the affix (but there are restrictions). Example:
When the dictionary file contains "drink/AS" then these words are possible:
drink drinks uses S suffix drinkable uses A suffix drinkables uses A suffix and then S suffix
Generally the flags of the suffix are added to the flags of the basic word, both are used for the word plus suffix. But the flags of the basic word are only used once for affixes, except that both one prefix and one suffix can be used when both support combining.
Specifically, the affix flags can be used for:
Suffixes on suffixes, as in the example above. This works once, thus you can have two suffixes on a word (plus one prefix).
Making the word with the affix rare, by using the spell-RARE flag.
Exclude the word with the affix from compounding, by using the spell-COMPOUNDFORBIDFLAG flag.
Allow the word with the affix to be part of a compound word on the side of the affix with the spell-COMPOUNDPERMITFLAG.
Use the NEEDCOMPOUND flag: word plus affix can only be used as part of a compound word. spell-NEEDCOMPOUND
Compound flags: word plus affix can be part of a compound word at the end, middle, start, etc. The flags are combined with the flags of the basic word. spell-compound
NEEDAFFIX: another affix is needed to make a valid word.
CIRCUMFIX, as explained just below.
Normally Vim gives an error for an extra field that does not start with '#'. This avoids errors going unnoticed. However, some files created for Myspell or Hunspell may contain many entries with an extra field. Use the IGNOREEXTRA flag to avoid lots of errors.
The CIRCUMFIX flag means a prefix and suffix must be added at the same time. If a prefix has the CIRCUMFIX flag then only suffixes with the CIRCUMFIX flag can be added, and the other way around. An alternative is to only specify the suffix, and give that suffix two flags: the required prefix and the NEEDAFFIX flag. spell-NEEDAFFIX
When an affix file has very many prefixes that apply to many words it's not possible to build the whole word list in memory. This applies to Hebrew (a list with all words is over a Gbyte). In that case applying prefixes must be postponed. This makes spell checking slower. It is indicated by this keyword in the .aff file:
Only prefixes without a chop string and without flags can be postponed. Prefixes with a chop string or with flags will still be included in the word list. An exception if the chop string is one character and equal to the last character of the added string, but in lower case. Thus when the chop string is used to allow the following word to start with an upper case letter.
The slash is used in the .dic file to separate the basic word from the affix letters and other flags. Unfortunately, this means you cannot use a slash in a word. Thus "TCP/IP" is not a word but "TCP" with the flags "IP". To include a slash in the word put a backslash before it: "TCP\/IP". In the rare case you want to use a backslash inside a word you need to use two backslashes. Any other use of the backslash is reserved for future expansion.
In the affix file a KEEPCASE line can be used to define the affix name used for keep-case words. Example:
This flag is not supported by Myspell. It has the meaning that case matters. This can be used if the word does not have the first letter in upper case at the start of a sentence. Example:
's morgens/= 's morgens 'S morgens 's Morgens 'S MORGENS 's Morgens 's Morgens 'S MORGENS 'S morgens 's morgens
The flag can also be used to avoid that the word matches when it is in all upper-case letters.
In the affix file a RARE line can be used to define the affix name used for rare words. Example:
Rare words are highlighted differently from bad words. This is to be used for words that are correct for the language, but are hardly ever used and could be a typing mistake anyway.
This flag can also be used on an affix, so that a basic word is not rare but the basic word plus affix is rare spell-affix-flags. However, if the word also appears as a good word in another way (e.g., in another region) it won't be marked as rare.
In the affix file a BAD line can be used to define the affix name used for bad words. Example:
This can be used to exclude words that would otherwise be good. For example "the the" in the .dic file:
Once a word has been marked as bad it won't be undone by encountering the same word as good.
The flag also applies to the word with affixes, thus this can be used to mark a whole bunch of related words as bad.
spell-FORBIDDENWORD FORBIDDENWORD can be used just like BAD. For compatibility with Hunspell.
spell-NEEDAFFIX The NEEDAFFIX flag is used to require that a word is used with an affix. The word itself is not a good word (unless there is an empty affix). Example:
A compound word is a longer word made by concatenating words that appear in the .dic file. To specify which words may be concatenated a character is used. This character is put in the list of affixes after the word. We will call this character a flag here. Obviously these flags must be different from any affix IDs used.
spell-COMPOUNDFLAG The Myspell compatible method uses one flag, specified with COMPOUNDFLAG. All words with this flag combine in any order. This means there is no control over which word comes first. Example:
spell-COMPOUNDRULE A more advanced method to specify how compound words can be formed uses multiple items with multiple flags. This is not compatible with Myspell 3.0. Let's start with an example:
The first line defines that words with the "c" flag can be concatenated in any order. The second line defines compound words that are made of one word with the "s" flag and one word with the "e" flag. With this dictionary:
You can make these words: bork borkbork borkborkbork (etc.) onion soup onionsoup
The COMPOUNDRULE item may appear multiple times. The argument is made out of one or more groups, where each group can be: one flag e.g., c alternate flags inside [] e.g., [abc] Optionally this may be followed by: * the group appears zero or more times, e.g., sm*e + the group appears one or more times, e.g., c+ ? the group appears zero times or once, e.g., x?
This is similar to the regexp pattern syntax (but not the same!). A few examples with the sequence of word flags they require: COMPOUNDRULE x+ x xx xxx etc. COMPOUNDRULE yz yz COMPOUNDRULE x+z xz xxz xxxz etc. COMPOUNDRULE yx+ yx yxx yxxx etc. COMPOUNDRULE xy?z xz xyz
COMPOUNDRULE [abc]z az bz cz COMPOUNDRULE [abc]+z az aaz abaz bz baz bcbz cz caz cbaz etc. COMPOUNDRULE a[xyz]+ ax axx axyz ay ayx ayzz az azy azxy etc. COMPOUNDRULE sm*e se sme smme smmme etc. COMPOUNDRULE s[xyz]*e se sxe sxye sxyxe sye syze sze szye szyxe etc.
A specific example: Allow a compound to be made of two words and a dash: In the .aff file: In the .dic file:
This allows for the word "start-end", but not "startend".
An additional implied rule is that, without further flags, a word with a prefix cannot be compounded after another word, and a word with a suffix cannot be compounded with a following word. Thus the affix cannot appear on the inside of a compound word. This can be changed with the spell-COMPOUNDPERMITFLAG.
spell-NEEDCOMPOUND The NEEDCOMPOUND flag is used to require that a word is used as part of a compound word. The word itself is not a good word. Example:
spell-ONLYINCOMPOUND The ONLYINCOMPOUND does exactly the same as NEEDCOMPOUND. Supported for compatibility with Hunspell.
spell-COMPOUNDMIN The minimal character length of a word used for compounding is specified with COMPOUNDMIN. Example:
When omitted there is no minimal length. Obviously you could just leave out the compound flag from short words instead, this feature is present for compatibility with Myspell.
spell-COMPOUNDWORDMAX The maximum number of words that can be concatenated into a compound word is specified with COMPOUNDWORDMAX. Example:
When omitted there is no maximum. It applies to all compound words.
To set a limit for words with specific flags make sure the items in COMPOUNDRULE where they appear don't allow too many words.
spell-COMPOUNDSYLMAX The maximum number of syllables that a compound word may contain is specified with COMPOUNDSYLMAX. Example:
This has no effect if there is no SYLLABLE item. Without COMPOUNDSYLMAX there is no limit on the number of syllables.
If both COMPOUNDWORDMAX and COMPOUNDSYLMAX are defined, a compound word is accepted if it fits one of the criteria, thus is either made from up to COMPOUNDWORDMAX words or contains up to COMPOUNDSYLMAX syllables.
spell-COMPOUNDFORBIDFLAG The COMPOUNDFORBIDFLAG specifies a flag that can be used on an affix. It means that the word plus affix cannot be used in a compound word. Example: affix file: dictionary:
This allows for "wordutil" and "wordutils" but not "wordutilize". Note: this doesn't work for postponed prefixes yet.
spell-COMPOUNDPERMITFLAG The COMPOUNDPERMITFLAG specifies a flag that can be used on an affix. It means that the word plus affix can also be used in a compound word in a way where the affix ends up halfway through the word. Without this flag that is not allowed. Note: this doesn't work for postponed prefixes yet.
spell-COMPOUNDROOT The COMPOUNDROOT flag is used for words in the dictionary that are already a compound. This means it counts for two words when checking the compounding rules. Can also be used for an affix to count the affix as a compounding word.
spell-CHECKCOMPOUNDPATTERN CHECKCOMPOUNDPATTERN is used to define patterns that, when matching at the position where two words are compounded together forbids the compound. For example:
This forbids compounding if the first word ends in "o" and the second word starts with "e".
The arguments must be plain text, no patterns are actually supported, despite the item name. Case is always ignored.
The Hunspell feature to use three arguments and flags is not supported.
spell-NOCOMPOUNDSUGS This item indicates that using compounding to make suggestions is not a good idea. Use this when compounding is used with very short or one-character words. E.g. to make numbers out of digits. Without this flag creating suggestions would spend most time trying all kind of weird compound words.
spell-SYLLABLE The SYLLABLE item defines characters or character sequences that are used to count the number of syllables in a word. Example:
Before the first slash is the set of characters that are counted for one syllable, also when repeated and mixed, until the next character that is not in this set. After the slash come sequences of characters that are counted for one syllable. These are preferred over using characters from the set. With the example "ideeen" has three syllables, counted by "i", "ee" and "e".
Only case-folded letters need to be included.
Another way to restrict compounding was mentioned above: Adding the spell-COMPOUNDFORBIDFLAG flag to an affix causes all words that are made with that affix to not be used for compounding.
For some languages, such as Thai, there is no space in between words. This looks like all words are compounded. To specify this use the NOBREAK item in the affix file, without arguments:
Vim will try to figure out where one word ends and a next starts. When there are spelling mistakes this may not be quite right.
spell-COMMON Common words can be specified with the COMMON item. This will give better suggestions when editing a short file. Example:
The words must be separated by white space, up to 25 per line. When multiple regions are specified in a ":mkspell" command the common words for all regions are combined and used for all regions.
spell-NOSPLITSUGS This item indicates that splitting a word to make suggestions is not a good idea. Split-word suggestions will appear only when there are few similar words.
spell-NOSUGGEST The flag specified with NOSUGGEST can be used for words that will not be suggested. Can be used for obscene words.
In the affix file REP items can be used to define common mistakes. This is used to make spelling suggestions. The items define the "from" text and the "to" replacement. Example:
The first line specifies the number of REP lines following. Vim ignores the number, but it must be there (for compatibility with Myspell).
Don't include simple one-character replacements or swaps. Vim will try these anyway. You can include whole words if you want to, but you might want to use the "file:" item in 'spellsuggest' instead.
You can include a space by using an underscore:
In the affix file MAP items can be used to define letters that are very much alike. This is mostly used for a letter with different accents. This is used to prefer suggestions with these letters substituted. Example:
The first line specifies the number of MAP lines following. Vim ignores the number, but the line must be there.
Each letter must appear in only one of the MAP items. It's a bit more efficient if the first letter is ASCII or at least one without accents.
When soundfolding is specified in the affix file then ":mkspell" will normally produce a .sug file next to the .spl file. This file is used to find suggestions by their sound-a-like form quickly. At the cost of a lot of memory (the amount depends on the number of words, :mkspell will display an estimate when it's done).
To avoid producing a .sug file use this item in the affix file:
Users can simply omit the .sug file if they don't want to use it.
In the affix file SAL items can be used to define the sounds-a-like mechanism to be used. The main items define the "from" text and the "to" replacement. Simplistic example:
There are a few rules and this can become quite complicated. An explanation how it works can be found in the Aspell manual: http://aspell.net/man-html/Phonetic-Code.html.
There are a few special items:
"1" has the same meaning as "true". Any other value means "false".
The SAL mechanism is complex and slow. A simpler mechanism is mapping all characters to another character, mapping similar sounding characters to the same character. At the same time this does case folding. You can not have both SAL items and simple soundfolding.
There are two items required: one to specify the characters that are mapped and one that specifies the characters they are mapped to. They must have exactly the same number of characters. Example:
In the example all vowels are mapped to the same character 'e'. Another method would be to leave out all vowels. Some characters that sound nearly the same and are often mixed up, such as 'm' and 'n', are mapped to the same character. Don't do this too much, all words will start looking alike.
Characters that do not appear in SOFOFROM will be left out, except that all white space is replaced by one space. Sequences of the same character in SOFOFROM are replaced by one.
You can use the soundfold() function to try out the results. Or set the 'verbose' option to see the score in the output of the z= command.
These items appear in the affix file of other spell checkers. In Vim they are ignored, not supported or defined in another way.
ACCENT (Hunspell) spell-ACCENT Use MAP instead. spell-MAP
BREAK (Hunspell) spell-BREAK Define break points. Unclear how it works exactly. Not supported.
CHECKCOMPOUNDCASE (Hunspell) spell-CHECKCOMPOUNDCASE Disallow uppercase letters at compound word boundaries. Not supported.
CHECKCOMPOUNDDUP (Hunspell) spell-CHECKCOMPOUNDDUP Disallow using the same word twice in a compound. Not supported.
CHECKCOMPOUNDREP (Hunspell) spell-CHECKCOMPOUNDREP Something about using REP items and compound words. Not supported.
CHECKCOMPOUNDTRIPLE (Hunspell) spell-CHECKCOMPOUNDTRIPLE Forbid three identical characters when compounding. Not supported.
CHECKSHARPS (Hunspell) spell-CHECKSHARPS SS letter pair in uppercased (German) words may be upper case sharp s (ß). Not supported.
COMPLEXPREFIXES (Hunspell) spell-COMPLEXPREFIXES Enables using two prefixes. Not supported.
COMPOUND (Hunspell) spell-COMPOUND This is one line with the count of COMPOUND items, followed by that many COMPOUND lines with a pattern. Remove the first line with the count and rename the other items to COMPOUNDRULE spell-COMPOUNDRULE
COMPOUNDBEGIN (Hunspell) spell-COMPOUNDBEGIN Words signed with COMPOUNDBEGIN may be first elements in compound words. Use COMPOUNDRULE instead. spell-COMPOUNDRULE
COMPOUNDLAST (Hunspell) spell-COMPOUNDLAST Words signed with COMPOUNDLAST may be last elements in compound words. Use COMPOUNDRULE instead. spell-COMPOUNDRULE
COMPOUNDEND (Hunspell) spell-COMPOUNDEND Probably the same as COMPOUNDLAST
COMPOUNDMIDDLE (Hunspell) spell-COMPOUNDMIDDLE Words signed with COMPOUNDMIDDLE may be middle elements in compound words. Use COMPOUNDRULE instead. spell-COMPOUNDRULE
COMPOUNDRULES (Hunspell) spell-COMPOUNDRULES Number of COMPOUNDRULE lines following. Ignored, but the argument must be a number.
KEY (Hunspell) spell-KEY Define characters that are close together on the keyboard. Used to give better suggestions. Not supported.
LANG (Hunspell) spell-LANG This specifies language-specific behavior. This actually moves part of the language knowledge into the program, therefore Vim does not support it. Each language property must be specified separately.
LEMMA_PRESENT (Hunspell) spell-LEMMA_PRESENT Only needed for morphological analysis.
MAXNGRAMSUGS (Hunspell) spell-MAXNGRAMSUGS Set number of n-gram suggestions. Not supported.
SUGSWITHDOTS (Hunspell) spell-SUGSWITHDOTS Adds dots to suggestions. Vim doesn't need this.
SYLLABLENUM (Hunspell) spell-SYLLABLENUM Not supported.
TRY (Myspell, Hunspell, others) spell-TRY Vim does not use the TRY item, it is ignored. For making suggestions the actual characters in the words are used, that is much more efficient.
WORDCHARS (Hunspell) spell-WORDCHARS Used to recognize words. Vim doesn't need it, because there is no need to separate words before checking them (using a trie instead of a hashtable).
When spell checking was going to be added to Vim a survey was done over the available spell checking libraries and programs. Unfortunately, the result was that none of them provided sufficient capabilities to be used as the spell checking engine in Vim, for various reasons:
Missing support for multi-byte encodings. At least UTF-8 must be supported, so that more than one language can be used in the same file. Doing on-the-fly conversion is not always possible (would require iconv support).
For the programs and libraries: Using them as-is would require installing them separately from Vim. That's mostly not impossible, but a drawback.
Performance: A few tests showed that it's possible to check spelling on the fly (while redrawing), just like syntax highlighting. But the mechanisms used by other code are much slower. Myspell uses a hashtable, for example. The affix compression that most spell checkers use makes it slower too.
For using an external program like aspell a communication mechanism would have to be setup. That's complicated to do in a portable way (Unix-only would be relatively simple, but that's not good enough). And performance will become a problem (lots of process switching involved).
Missing support for words with non-word characters, such as "Etten-Leur" and "et al.", would require marking the pieces of them OK, lowering the reliability.
Missing support for regions or dialects. Makes it difficult to accept all English words and highlight non-Canadian words differently.
Missing support for rare words. Many words are correct but hardly ever used and could be a misspelled often-used word.
For making suggestions the speed is less important and requiring to install another program or library would be acceptable. But the word lists probably differ, the suggestions may be wrong words.
Spelling suggestions develop-spell-suggestions
For making suggestions there are two basic mechanisms: 1. Try changing the bad word a little bit and check for a match with a good word. Or go through the list of good words, change them a little bit and check for a match with the bad word. The changes are deleting a character, inserting a character, swapping two characters, etc. 2. Perform soundfolding on both the bad word and the good words and then find matches, possibly with a few changes like with the first mechanism.
The first is good for finding typing mistakes. After experimenting with hashtables and looking at solutions from other spell checkers the conclusion was that a trie (a kind of tree structure) is ideal for this. Both for reducing memory use and being able to try sensible changes. For example, when inserting a character only characters that lead to good words need to be tried. Other mechanisms (with hashtables) need to try all possible letters at every position in the word. Also, a hashtable has the requirement that word boundaries are identified separately, while a trie does not require this. That makes the mechanism a lot simpler.
Soundfolding is useful when someone knows how the words sounds but doesn't know how it is spelled. For example, the word "dictionary" might be written as "daktonerie". The number of changes that the first method would need to try is very big, it's hard to find the good word that way. After soundfolding the words become "tktnr" and "tkxnry", these differ by only two letters.
To find words by their soundfolded equivalent (soundalike word) we need a list of all soundfolded words. A few experiments have been done to find out what the best method is. Alternatives: 1. Do the sound folding on the fly when looking for suggestions. This means walking through the trie of good words, soundfolding each word and checking how different it is from the bad word. This is very efficient for memory use, but takes a long time. On a fast PC it takes a couple of seconds for English, which can be acceptable for interactive use. But for some languages it takes more than ten seconds (e.g., German, Catalan), which is unacceptable slow. For batch processing (automatic corrections) it's too slow for all languages. 2. Use a trie for the soundfolded words, so that searching can be done just like how it works without soundfolding. This requires remembering a list of good words for each soundfolded word. This makes finding matches very fast but requires quite a lot of memory, in the order of 1 to 10 Mbyte. For some languages more than the original word list. 3. Like the second alternative, but reduce the amount of memory by using affix compression and store only the soundfolded basic word. This is what Aspell does. Disadvantage is that affixes need to be stripped from the bad word before soundfolding it, which means that mistakes at the start and/or end of the word will cause the mechanism to fail. Also, this becomes slow when the bad word is quite different from the good word.
The choice made is to use the second mechanism and use a separate file. This way a user with sufficient memory can get very good suggestions while a user who is short of memory or just wants the spell checking and no suggestions doesn't use so much memory.
Word frequency
For sorting suggestions it helps to know which words are common. In theory we could store a word frequency with the word in the dictionary. However, this requires storing a count per word. That degrades word tree compression a lot. And maintaining the word frequency for all languages will be a heavy task. Also, it would be nice to prefer words that are already in the text. This way the words that appear in the specific text are preferred for suggestions.
What has been implemented is to count words that have been seen during displaying. A hashtable is used to quickly find the word count. The count is initialized from words listed in COMMON items in the affix file, so that it also works when starting a new file.
This isn't ideal, because the longer Vim is running the higher the counts become. But in practice it is a noticeable improvement over not using the word count.
Commands index
Quick reference