Major (and minor) modes in Vile

This document was originally written in May, 1998, and has been updated periodically to reflect the proposed implementation of and use of major modes in vile.

My goal was to extend the notion of the C mode (cmode) to allow runtime definable major modes.

Originally, vile supported a C mode that included a collection of modes useful for editing C program source:

as well as this builtin functionality:

Both the modes and functionality are extensions of other features in vile. It would be useful to combine modes to support other languages in a similar fashion. Likewise, the autoindention, etc., could be parameterized and made reusable to support other languages. For an initial implementation, I focused on the combining of modes, providing a structure for the parameterization.

One thing that was not clear to many users was the manner in which the C mode was attached to a buffer. It was set as a boolean - if active before a buffer was loaded, then vile checked the file suffix to see if it matched the c-suffixes mode, and if so, set the C mode for the buffer. C mode could also be explicitly set by a ":setl cmode", and unset by ":setl nocmode". In the new scheme,


These are the commands which I originally thought necessary:

The {majormode} is a new symbol.

The {minormode} can be any one of the existing buffer modes, except for a {majormode}. To make name-completion simple, we use the term 'submode'.

Later, I added features to make majormodes simpler to configure:


      define-majormode c
      ; Declares a mode 'c', and corresponding symbol 'cmode'

      define-submode c suffixes="\\.\\(\\([Cchisyl]\\)\\|CC\\|cc|cpp\\|cxx\\|hxx\\|scm\\)$"
      ; Specifies the filename suffixes which control whether a newly-loaded
      ; buffer is set to 'c' mode.

      define-submode c tabstop=4
      define-submode c shiftwidth=4
      ; Defines the 'c' tabstop and shiftwidth.  If no "define-submode"
      ; command is given, no separate symbol is defined.

As an example, to define a new major mode for perl programming, you might include the following in your .vilerc file:

      define-majormode perl
      define-submode perl preamble "^#.*perl\\>"
      define-submode perl suffixes '\.\(pm\|t\)$'
      define-submode perl shiftwidth 4

To avoid the tediousness of this syntax, use the ~with and ~endwidth keywords, like so:

      define-majormode perl
      ~with define-submode perl
          preamble "^#.*perl\\>"
          suffixes '\.\(pm\|t\)$'
          shiftwidth 4

You can define several regions using ~with and (after the first) ~elsewith blocks. Each block specifies a set of tokens which are substituted at the beginning of each line. Each ~elsewith block should have a group identifier to distinguish it from the others. The reason for providing multiple blocks is to implement complex fence-matching for different language features.

For example, in the definition of cshmode, the symbols "fence-XXX" give patterns which vile can use to move the cursor from one if/elif/else/fi marker to the next as you press "%". The other settings such as "suf" in the first block are settings that apply to the majormode itself:

      define-mode csh
      ~with define-submode csh
          suf '\.\(csh[^/]*\|login\|logout\)$'
          pre '^#!\s*\/.*csh\>\(\s*-[a-z]\+\)*\s*$'
          filtername  'vile-sh-filt -k csh'
          comment-prefix      '^\s*#'
          comments    '^\s*#\s*$'
          fence-if    '^\s*\<if\>.*\<then\>'
          fence-elif  '^\s*\<else\s*if\>'
          fence-else  '^\s*\<else\>'
          fence-fi    '^\s*\<endif\>'
      ~elsewith define-submode csh group 'case'
          fence-if    '^\s*\<switch\>\s*(.*)'
          fence-elif  '^\s*\<case\>.*:'
          fence-else  '^\s*\<default\>\s*:'
          fence-fi    '^\s*\<endsw\>'
      ~elsewith define-submode csh group 'loop'
          fence-if    '^\s*\<foreach\s\+.*\|while\>\s*(.*)'
          fence-fi    '^\s*\<end\>'

The "define-majormode" command

This takes a single argument, a majormode name. To follow existing convention, the string "mode" is automatically appended to the given name. Associated modes are defined or modified with the define-submode command. Vile maintains a list of majormodes. Only one majormode can be associated with a buffer (none need be associated). After definition, a majormode can be set or unset just like any other buffer mode:

      define-majormode c
      ; defines "cmode"

      setl cmode
      ; sets the mode for the current buffer

      setl nocmode
      ; clear c mode (existing implementation)

      unsetl cmode
      ; clear c mode

The restriction to a single majormode is because mode values are internally represented as structures with a pointer. The pointer denotes which value (currently local or global) is used. The majormode implementation adds a level to this, e.g.,
        value -> self (local)
        value -> global (global)
        value -> major (majormode)

When a majormode is defined, an array of the existing minor mode values is allocated, all pointing to the global modes. The define-submode command modifies these to make them local pointers. When a buffer is associated with a majormode, all of its buffer mode values are pointed to the majormode's values. (To keep the bookkeeping straight, modifying a global buffer mode must also modify the copies of non-local buffer mode values).

The "derive-majormode" command

This is used to clone an existing majormode, using a new name. The command takes two parameters:

  1. the new majormode name
  2. the old majormode name, used as a source of settings

If the new majormode already exists, it is not removed. Instead, settings are copied from the old majormode into the existing majormode. In principle, a series of these commands could be used to merge several different majormodes.

The "define-submode" command

This command sets local values of buffer modes for the given majormode, e.g.,

      define-submode c autoindent

The majormode name is required. The names after the majormode name are names of buffer modes with a corresponding value. Any number of modes can be specified in a single command, e.g.,

      define-submode c autoindent tabstop=4

For each mode given, vile defines corresponding names by which they can be referenced, e.g., "c-autoindent" (from the long mode name "autoindent") and "cai" (from the short mode name "ai" for autoindent). The long mode name is always appended with a hyphen, and the short mode name is appended without a hyphen.

The term "submode" is used in the command rather than the more natural "minor mode" to simplify name-completion.

The following are keywords that aren't minor modes, but are recognized solely by the define-submode command:

The full pathname which is tested to trigger association with a majormode, e.g., "RCS/*,v".
The filename which is tested to trigger association with a majormode, e.g., "Makefile".
The filename suffix which is tested to trigger association with a majormode (e.g., c-suffixes) Note that since the default value for the c-suffixes mode is a regular expression that will match some other file types (C++ files, for instance), if you define a new major mode for one of those suffixes you may want to reset c-suffixes to something less inclusive.
Regular expression, matched against the beginning of a file (the first line) used to trigger association with a majormode (e.g., "^!#.*\/perl[^a-z]").
A shell command telling which filter to call, and how to do that. The form of the command is limited, since it is interpreted for internal calls:
        {filter} {options}

where {filter} is the filename for the filter, e.g.,


and options include:

-k mode
The keyword filename, omitting ".keywords". For instance, the C filter vile-c-filt is used for C, C++, JavaScript and Java by changing the keyword file.
Quit after loading the keyword definitions.
-t tabs
Set the tabstops, used in the imake filter to check for coincidental matches between a tab and spaces that are the same number of columns.
Debug traces, mainly to show which keyword files are loaded, and what information is parsed from them.

Other options may be implemented that are specific to a filter program. For instance, vile-c-filt recognizes a -p option to mark preprocessor lines with an error (used for Java).

A string that tells vile how to rank the majormode. Normally vile checks each mode in order by name, but there are special cases where you may want to check one mode before another. For example, the C++ majormode cppmode would be checked after cmode, but that uses suffixes which are a subset of the built-in cmode's suffixes and would not be found. So cppmode is qualified before="c".
followed by a name, defines an alternative set of submode values.

Currently used only for complex fences, this could be applied to simple fences, and (with new flags not yet defined) extend both styles of fences for indentation and formatting.

These are "complex" fences, which are matched one expression per line. The names are "fence-" followed by any of the following keywords with a regular expression:
        if, elif, else, fi

Vile searches through all groups of complex fences for a match before trying simple fences.

These are "simple" fences, which can be matched any number of times per line. The pairs need not appear on the same line. Like complex fences, they have an implied order. The default values support C-style comments, with the '%' going between "/*" and "*/". The names use "fence-" followed any of the following keywords with a regular expression: begin, end.
indent (not implemented)
The keyword "cstyle", or any of the following keywords with a regular expression: begin, end, if, then, else, elif, endif.

Other features which should be controlled by majormodes include limiting the scope of the entab and detab commands.

The "remove-majormode" command

This command has two forms:

remove-majormode {majormode}
This removes the definition of the majormode. Buffers that were associated with the mode revert to no majormode.


remove-majormode {majormode} {name}
This removes the value of {name} from {majormode}, leaving it set to the global value, if any.

The "remove-submode" command

Remove the special association of a submode from a majormode.


The original builtin C/C++ majormode description is equivalent to

      define-mode c
      ~with define-submode c
              suffix "\\.\\(\\([Cchisyl]\\)\\|CC\\|cc|cpp\\|cxx\\|hxx\\|scm\\)$"

Note that the following are equivalent once you have defined the majormode "c":

      set cts=8
      set c-tabstop=8
      define-submode c tabstop=8


Most of this was written by Thomas Dickey, with fixes from Clark Morgan and Steven Lembark.