You are looking at the top node of the Info tree documenting SML-MODE (Version 3.3). Not all functions are documented here, but those that aren't you probably won't miss. All commands and settable variables have built-in documentation, as per usual Emacs conventions.
You can freely copy, modify and redistribute SML mode because it's made available under the liberal terms of the GNU General Public License.
GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version.
SML mode is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with GNU Emacs; see the file COPYING. If not, write to the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
SML mode is a major mode for Emacs for editing Standard ML. It has some novel bugs, and some nice features:
| automatically determines if it is used
in a case or fun construct, and indents the next line as appropriate,
inserting => or the name of the function.
The distribution contains several Emacs Lisp files--this is for ease of maintenance, you can concatenate them if you're careful:
There is also the Texinfo generated info file:
With luck your system administrator will have installed SML mode somewhere convenient, so all you have to do is put the line
(require 'sml-site)
in your `.emacs' configuration file and all will be well--you can skip the rest of this getting started section. Otherwise you will need to tell Emacs where to find all the SML mode `.el' files, and when to use them. The where is addressed by locating the Lisp code on your Emacs Lisp load path--you may have to create a directory for this, say `/home/mjm/elisp', and then insert the following lines in your `/home/mjm/.emacs' file(1):
(setq load-path (cons "/home/mjm/elisp" load-path)) (autoload 'sml-mode "sml-mode" "Major mode for editing SML." t)
The first line adjusts Emacs' internal search path so it can locate the Lisp source you have copied to that directory; the second line tells Emacs to load the code automatically when it is needed. You can then switch any Emacs buffer into SML mode by entering the command
M-x sml-mode
It is usually more convenient to have Emacs automatically place the buffer in SML mode whenever you visit a file containing ML programs. The simplest way of achieving this is to put something like
(setq auto-mode-alist
(append '(("\\.sml$" . sml-mode)
("\\.sig$" . sml-mode)
("\\.ML$" . sml-mode)) auto-mode-alist))
also in your `.emacs' file. Subsequently (after a restart), any files with these extensions will be placed in SML mode buffers when you visit them.
You may want to pre-compile the `sml-*.el' files (M-x byte-compile-file) for greater speed--byte compiled code loads and runs somewhat faster.
You're reading it. Apart from the on-line info tree (C-h i is the
Emacs key to enter the info system--you should follow the brief
tutorial if this is unfamiliar), there are further details on specific
commands in their documentation strings. Only the most useful
SML mode commands are documented in the info tree: to find out more
use Emacs' help facilities.
Briefly, to get help on a specific function use C-h f and enter
the command name. All (almost all, then) SML mode commands begin
with sml-, so if you type this and press TAB (for
completion) you will get a list of all commands. Another way is to use
C-h a and enter the string sml. This is command apropos; it
will list all commands with that sub-string in their names, and any key
binding they may have in the current buffer. Command apropos gives a
one-line synopsis of what each command does.
Some commands are also variables--such things are allowed in Lisp, if not in ML! See section Command Index, for a list of (info) documented functions. See section Variable Index, for a list of user settable variables to control the behaviour of SML mode.
Before accessing this information on-line from within Emacs you may have
to set the variable sml-mode-info. Put in your `.emacs' file
something like:
(setq sml-mode-info "/home/mjm/info/sml-mode.info")
When different from the default this variable should be a string giving the absolute name of the `.info' file. Then C-c C-i in SML mode (i.e., the command M-x sml-mode-info) will bring up the manual. This help is also accessible from the menu. (Resetting this variable will not be necessary if your site administrator has been kind enough to install SML mode and its attendant documentation in the Emacs hierarchy.)
Now SML mode provides just a few additional editing commands. Most of the work (see section Credit & Blame) has gone into implementing the indentation algorithm which, if you think about it, has to be complicated for a language like ML. See section Indentation defaults, for details on how to control some of the behaviour of the indentation algorithm. Principal goodies are the `electric pipe' feature, and the ability to insert common SML forms (macros or templates).
Emacs is all hooks of course. A hook is a variable: if the variable is non-nil it binds a list of Emacs Lisp functions to be run in some order (usually left to right). You can customise SML mode with these hooks:
nil
This is run every time a new SML mode buffer is created (or if you type M-x sml-mode). This is one place to put your preferred key bindings. See section Configuration Summary, for some examples.
'sml-mode-version
Another, maybe better, place for key bindings. This hook is only run when SML mode is loaded into Emacs. See section Configuration Summary.
sml-load-hook so you can easily
tell which version of SML mode you are running.
ML is a complicated language to parse, let alone compile. The indentation algorithm is a little wooden (for some tastes), and the best advice is not to fight it! There are several variables that can be adjusted to control the indentation algorithm (see section Indentation defaults, below).
This command indents the current line. If you set the indentation of the
previous line by hand, sml-indent-line will indent relative to
this setting.
Indent the current region. Be patient if the region is large (like the whole buffer).
Unindents the line to the next outer level of indentation.
Further indentation commands that Emacs provides (generically, for all modes) that you may like to recall:
As with other language modes, M-; gives you a comment at the end
of the current line. The column where the comment starts is determined
by the variable comment-column---default is 40, but it can be
changed with set-comment-column (on C-x ; by default).
Electric keys are generally pretty irritating, so those provided by SML mode are fairly muted. The only truly electric key is ;, and this has to be enabled to take effect.
When the point is in a `case' statement this opens a new line, indents
and inserts | => leaving point just before the double arrow; if
the enclosing construct is a `fun' declaration, the newline is indented
and the function name copied at the appropriate column. Generally, try
it whenever a | is wanted--you'll like it!
Just inserts a semi-colon, usually. The behaviour of this command is
governed by the variable sml-electric-semi-mode.
nil
If this variable is nil, sml-electric-semi just inserts a
semi-colon, otherwise it inserts a semi-colon and a newline, and indents
the newline for SML. The command toggles the value of the variable; if
you give the command a prefix argument (i.e., C-u M-x
sml-electric-semi-mode) this always disables the electric effect of
;.
Interactive short-cut to insert common ML forms (a.k.a. macros, or templates). Recognised forms are `let', `local', `case', `abstype', `datatype', `signature', `structure', and `functor'. Except for `let' and `local', these will prompt for appropriate parameters like functor name and signature, etc.. This command prompts in the mini-buffer, with completion.
By default C-c RET will insert at point, with the indentation of the current column; if you give a prefix argument (i.e., C-u C-c RET) the command will insert a newline first, indent, and then insert the template.
sml-insert-form is also extensible: see section Configuration Summary for
further details.
Several variables try to control the indentation algorithm and other features of SML mode. For these user settable variables there is generally a function of the same name that does the job--look for them in the menu under Format/Mode Variables.
4
This variable controls the block indentation level. The command prompts for a numeric value unless a numeric prefix is provided instead. For example M-2 M-x sml-indent-level will set the variable to 2 without prompting.
-2
This variable adjusts the indentation level for lines that begin with a
| (after any white space). The extra offset is usually negative.
The command prompts for a numeric value unless a numeric prefix is
provided instead.
1000
The number of characters the indentation algorithm searches for an
opening parenthesis. 1000 characters is about 30-40 lines; larger values
mean slower indentation. If the value of the variable is nil this
means the indentation algorithm won't look back at all.
If the default values are not acceptable you can set these variables permanently in your `.emacs' file. See section Configuration Summary, for details and examples. Three further variables control the behaviour of indentation.
nil
How to indent `case' expressions:
Ift: Ifnil: case expr case expr of of exp1 => ... exp1 => ... | exp2 => ... | exp2 => ...
The first seems to be the standard in SML/NJ. The second is the (nicer?) default.
nil
Nested `if-then-else' expressions have the following indentation depending on the value.
Ift: Ifnil: if exp1 then exp2 if exp1 then exp2 else if exp3 then exp4 else if exp3 then exp4 else if exp5 then exp6 else if exp5 then exp6 else exp7 else exp7
t
Determines how to indent `let', `struct', etc..
Ift: Ifnil: fun foo bar = let fun foo bar = let val p = 4 val p = 4 in in bar + p bar + p end end
sml-type-of-indent will not have any effect if the starting
keyword is the first word on the line.
The most useful feature of SML mode is that it provides a convenient interface to the compiler. How serious users of ML put up with a teletype interface to the compiler is beyond me... but perhaps there are other interfaces to compilers that require one to part with serious money. Such remarks can quickly become dated--in this case, let's hope so!
Anyway, SML mode provides an interaction mode,
inferior-sml-mode, where the compiler runs in a separate buffer
in a window or frame of its own. You can use this buffer just like a
terminal, but it's usually more convenient to mark some text in the
SML mode buffer and have Emacs communicate with the sub-process. The
features discussed below are syntax-independent, so they should work
with a wide range of ML-like tools and compilers. See section Process defaults, for some hints.
inferior-sml-mode is a specialisation of the `comint'
package that comes with GNU Emacs and GNU XEmacs.
Start your favourite ML compiler with the command
M-x sml
This creates a process interaction buffer that inherits some key bindings from SML mode and from `comint' (see section `Shell Mode' in The Emacs Editor Manual). Starting the ML compiler adds some functions to SML mode buffers so that program text can be communicated between editor and compiler (see section Speaking to the compiler).
The name of the ML compiler is the first thing you should know how to specify:
"sml"
The program to run as ML. You might need to specify the full path name of the program.
""
Useful for Poly/ML users who may supply a database file, or others who
have wrappers for setting various options around the command to run the
compiler. Moscow ML people might set this to "-P full", etc..
The variable sml-program-name is a string holding the name
of the program as you would type it at the shell. You
can always choose a program different to the default by invoking
C-u M-x sml
With the prefix argument Emacs will prompt for the command name and any command line arguments to pass to the compiler. Thereafter Emacs will use this new name as the default, but for a permanent change you should set this in your `.emacs' with, e.g.:
(setq sml-program-name "nj-sml")
You probably shouldn't set this in sml-mode-hook because that
will interfere if you occasionally run a different compiler (e.g.,
poly or hol90).
nil
M-x sml runs comint-mode-hook and
inferior-sml-mode-hook hooks in that order, but after the
compiler is started. Use inferior-sml-mode-hook to set any
comint buffer-local configurations for SML mode you like.
nil
This hook is analogous to sml-load-hook and is run just after the
code for inferior-sml-mode is loaded into Emacs. Use this to set
process defaults, and preferred key bindings for the interaction buffer.
Switch from the SML buffer to the interaction buffer. By default point will be placed at the end of the process buffer, but a prefix argument will leave point wherever it was before. If you try C-c C-s before an ML process has been started, you'll just get an error message to the effect that there's no current process buffer.
(if window-system t nil)
If t this indicates to switch-to-sml and other functions
that the interaction buffer where ML is running will be displayed on its
own, dedicated frame; otherwise the interaction buffer will appear on
the current frame, splitting the window if necessary. The default means
SML mode will try and use a dedicated frame if you are running Emacs
under X Windows (say), but not otherwise. The variable
sml-display-frame-alist configures the dedicated frame's
appearance (C-h v sml-display-frame-alist for details).
sml-cd-command specifies the compiler command to invoke
(see section Process defaults).
Several commands are defined for sending program fragments to the running compiler. Each of the following commands takes a prefix argument that will switch the input focus to the process buffer afterwards (leaving point at the end of the buffer):
Send a `use file' command to the current ML process. The variable
sml-use-command is used to define the correct template for the
command to invoke (see section Process defaults). The default file is the
file associated with the current buffer, or the last file loaded if you
are in the interaction buffer.
Send the current region of text in the SML buffer.
sml-send-region-and-go is a similar command for you to bind in
SML mode if you wish: it'll send the region and then switch-to-sml.
It's sometimes irritating to do all that C-@ and C-c C-r
stuff to send regions to the ML process, so if you are running Emacs
under X Windows (say) you can do the same job by holding down both the
SHIFT and META keys, and dragging with mouse button one over
the region. This will temporarily highlight the region as you move the
mouse, like mouse-drag-region (i.e., down-mouse-1),
and send the highlighted text straight into the jaws of the ML compiler.
If you only click the mouse button, instead of dragging, the region of text sent to the compiler is delimited by the current position of point and the place where you click the mouse. In neither case will the command set the region.
Send the contents of the current buffer to ML.
By and large, Emacs can nowadays quite happily send large chunks of text
to its subprocesses (`comint' does input splitting). However, it is
still probably safest(2) to send
larger program fragments to ML via the temporary file mechanism. This,
for sml-send-region and other SML mode commands that use it
in some way, takes advantage of the ML compiler's ability to open a file
and compile the contents by making a temporary file of the indicated
text. Two variables of interest are:
0
Determines what constitutes a large program fragment. A value of 512,
say, will declare half a kilobyte a suitable threshold and larger
fragments will be sent via a temporary file. A value of 0 means
all text is sent via a temporary file; the value nil
inhibits the temporary file mechanism altogether.
(make-temp-name "/tmp/ml")
A string that gives the name of the temporary file to use. This default ensures Emacs will invent a unique name for this purpose for use throughout the rest of the editing session. Only one temporary file is used.
Another reason, you might well say the reason, for using the temporary file mechanism is that error messages reported by the ML compiler (see section Finding errors) are generally useless to SML mode unless a real file is associated with the input (an embedded use file will count as a real file). Of course, this all rather depends on the compiler producing sensible error messages, and on SML mode being able to parse them.
SML mode provides one customisable function for locating the source
position of errors reported by the compiler. This should work whether
you type use "puzzle.sml"; into the interaction buffer, or use
one of the mechanisms provided for sending programs directly to the
compiler---see section Speaking to the compiler.
Jump to the source location of the next error reported by the compiler.
If the function bound to sml-error-parser returns a range of
character positions for the location of the error in the source file,
sml-next-error will put the mark at the end of the range with
point at the beginning; it may also highlight the region specified; it
will also echo the one-line text of the error message if the error
parser returns one.(3)
If you enter C-u C-c` instead, the command (a.k.a.
sml-skip-errors) skips past all the remaining error messages and
removes any error overlay in the current buffer. Note that C-c`
also works in the ML interaction buffer (by default).
t
Legal default values for this buffer-local variable are t and
nil. The variable attains a local value in each SML mode
buffer when the default is t; in this case the local value is an
overlay (or extent in XEmacs speak), and this means
sml-next-error will highlight errors in the buffer when it can.
If the default is nil it stays that way and sml-next-error
will not highlight anything, ever.
The command M-x sml-error-overlay will set the overlay around the current region, or remove the overlay if a prefix argument is given (i.e., C-u M-x sml-error-overlay removes the overlay, but this functionality can be accessed from the menu to save typing).
Note that SML mode will usually locate errors relative to the start
of the last major program fragment sent to the compiler (via
sml-load-file, etc.), but if you don't use the temporary file
mechanism to communicate text to the ML process (see section Process defaults), errors will generally not be located at all.
The process interaction code is independent of the compiler used, deliberately, so SML mode will work with a variety of ML compilers and ML-based tools. There are therefore a number of variables that may need to be set correctly before SML mode can speak to the compiler. Things are by default set up for Standard ML of New Jersey, but switching to a new system is quite easy--very easy if you are using Poly/ML or Moscow ML as these are supported by libraries bundled with SML mode.
"use \"%s\""
Use file command template. Emacs will replace the %s with a file
name. Note that Emacs requires double quote characters inside strings
to be quoted with a backslash.
"OS.FileSys.chDir \"%s\""
Compiler command to change the working directory. Not all ML systems support this feature (well, Edinburgh (core) ML didn't), but they should.
"^[\-=] *"
Matches the ML compiler's prompt: `comint' uses this for various purposes.
To customise error reportage for different ML compilers you need to set
two further variables before sml-next-error can be useful:
sml-smlnj-error-regexp
This is the regular expression for matching the start of an error message. The default matches the Standard ML of New Jersey compiler's Error and Warning messages. If you don't want stop at Warnings try, for example:
"^[-= ]*.+:[0-9]+\\.[0-9]+.+Error:"
If you're using Edinburgh (core) ML try "^Parse error:".
'sml-smlnj-error-parser
The function that actually parses the error message. Again, the default is for SML/NJ. If you need to change this you may have to do a little Emacs Lisp programming.
Note that bundled libraries supply an sml-mosml-error-parser and
an sml-poly-ml-error-parser, and set all the attendant compiler
variables. See section Advanced Topics, for tips on how to program your own
compiler extension to SML mode.
This (sort of pedagogic) section gives more information on how to configure SML mode: menus, key bindings, hooks and highlighting are discussed, along with a few other random topics. First, though, the auxiliary files `sml-poly-ml.el' and `sml-mosml.el' define defaults for these popular (?) ML compilers--Poly/ML and Moscow ML, respectively. One way to setup SML mode to use Moscow ML is to add to your `.emacs':
(defun my-mosml-setup () "Initialise inferior SML mode for Moscow ML." (load-library "sml-mosml.el") (setq sml-program-name "/home/mjm/mosml/bin/mosml")) (add-hook 'inferior-sml-load-hook 'my-mosml-setup)
which creates a hook function my-mosml-setup and adds it to
inferior-sml-load-hook so that the defaults for
sml-error-regexp and its ilk (see section Process defaults) are
correctly initialised; I have to set sml-program-name explicitly
here because that directory isn't on my (Unix) PATH. The story is
similar if you use Poly/ML.
Note, by the way, that order matters here: the load-library call
comes first because the default for sml-program-name in
`sml-mosml.el' is just "mosml".
The auxiliary libraries bundled with SML mode define commands
sml-mosml and sml-poly-ml (there's also an
sml-smlnj for uniformity); these commands prompt for suitable
values for sml-program-name and sml-default-arg before
starting the compiler and setting the other process defaults. A prefix
argument will give you the builtin defaults with no questions asked.
One way to set SML mode variables (see section Indentation defaults), and other defaults, is through the
sml-mode-hook in your `.emacs'. A simple example:
(defun my-sml-mode-hook () "Local defaults for SML mode" (setq sml-indent-level 2) ; conserve on horizontal space (setq words-include-escape t) ; \ loses word break status (setq indent-tabs-mode nil)) ; never ever indent with tabs (add-hook 'sml-mode-hook 'my-sml-mode-hook)
The body of my-sml-mode-hook is a sequence of bindings. In this
case it is not really necessary to set sml-indent-level in a hook
because this variable is global (most SML mode variables are). With
similar effect:
(setq sml-indent-level 2)
anywhere in your `.emacs' file (but probably on
sml-load-hook). The variable indent-tabs-mode is
automatically made local to the current buffer whenever it is set
explicitly, so it must be set in a hook if you always want
SML mode to behave like this. The same goes for the buffer-local
sml-error-overlay; since this is globally t by default,
set it globally nil if you never want errors highlighted:
(setq-default sml-error-overlay nil)
Again, on sml-load-hook would probably be the best place.
Another hook is inferior-sml-mode-hook. This can be used to
control the behaviour of the interaction buffer through various
variables meaningful to `comint'-based packages:
(defun my-inf-sml-mode-hook () "Local defaults for inferior SML mode" (add-hook 'comint-output-filter-functions 'comint-truncate-buffer) (setq comint-scroll-show-maximum-output t) (setq comint-input-autoexpand nil)) (add-hook 'inferior-sml-mode-hook 'my-inf-sml-mode-hook)
Again, the body is a sequence of bindings. Unless you run several ML
compilers simultaneously under one Emacs, this hook will normally only
get run once. You might want to look up the documentation (C-h v
and C-h f) for these buffer-local comint things.
Customisation (in Emacs) usually entails putting favourite commands on
easily remembered keys. Two `keymaps' are defined in SML mode: one
is effective in program text buffers (sml-mode-map) and the other
is effective in interaction buffers (inferior-sml-mode-map).
The initial design ensures that (many of) the default key bindings from
the former keymap will also be available in the latter (e.g.,
C-c`).
Type C-h m in an SML mode buffer to find the default key bindings (and similarly in an ML interaction buffer), and use the hooks provided to install your preferred key bindings. Given that the keymaps are global (variables):
(defun my-sml-load-hook () "Global defaults for SML mode" (define-key sml-mode-map "\C-cd" 'sml-cd) (define-key sml-mode-map "\C-co" 'sml-error-overlay)) (add-hook 'sml-load-hook 'my-sml-load-hook)
This has the effect of binding sml-cd to the key C-c d, and
the command sml-error-overlay to the key C-c o. If you want
the same behaviour from C-c d in the ML buffer:
(defun my-inf-sml-load-hook () "Global defaults for inferior SML mode" (define-key inferior-sml-mode-map "\C-cd" 'sml-cd) ;; NB. for SML/NJ '96 (setq sml-cd-command "OS.FileSys.chDir \"%s\"")) (add-hook 'inferior-sml-load-hook 'my-inf-sml-load-hook)
There is nothing to stop you rebuilding the entire keymap for
SML mode and the ML interaction buffer in your `.emacs' of
course: SML mode won't define sml-mode-map or
inferior-sml-mode-map if you have already done so.
Menus are useful for fiddling with mode defaults and finding out what
keys commands are on if you are forgetful (not all commands are listed
in the menu). For menus to appear in the menu bar under GNU Emacs or GNU
XEmacs, the editor must be able to find one of two packages--i.e., one
or both must be on your load-path. The first option is
`easymenu' which is distributed with GNU Emacs. Easy!
The second option is `auc-menu' which was written by Per Abrahamsen and distributed with AUCTeX, but it is independently available from the IESD lisp archive(4)} at Aalborg. You'll also find `auc-menu' is available from the LCD archive(5)}, the main repository for all Emacs Lisp. The advantage of `auc-menu' is that it works with XEmacs too.
Notice that certain menu entries are not illuminated at first--these are generally functions that depend on there being an ML process running with which to communicate.
Highlighting is very handy for picking out keywords in the program text, spotting misspelled kewyords, and, if you have Emacs' `ps-print' package installed (you usually do these days), obtaining pretty, even colourful code listings--quite properly for your colourful ML programs.
Various highlight (hilite, if you spell real bad!) packages are available for GNU Emacs 19, and GNU XEmacs. SML mode can use either `hilit19' which only comes with Emacs, or `font-lock' which is the package of choice with XEmacs. If you are not familiar with these highlight packages you'll have to check their sources for installation guidelines, etc..
Use sml-load-hook to tell Emacs which scheme you prefer for
SML mode. For example:
(add-hook 'sml-load-hook '(lambda () (require 'sml-font)))
This ensures the SML extensions to `font-lock' will be available
once SML mode loads (from `sml-font.el'---if you prefer
`hilit19' you should (require 'sml-hilite) instead.
The variable sml-font-lock-extra-keywords is for further
customising `font-lock' for SML mode. The value of the variable
should be a list of strings, each of which is a regular expression that
should match the desired keyword exactly. Here's an example:
(setq sml-font-lock-extra-keywords
'("\\babstraction\\b" "\\bfunsig\\b" "=>" "::"))))
The \b marks a word boundary, according to the syntax table
defined for SML mode. Backslash must be quoted inside a string.
See section `Regexps' in The Emacs Editor Manual, for a summary of Emacs'
regular expression syntax.
Finally, the variable sml-font-lock-auto-on can be used to
control whether or not `font-lock' should be enabled by default in
SML mode buffers; it is enabled by default. The sml-hilite
package is customisable, but only with regard to colour changes.
These forms are bloody useless; can't we have better ones?
You can indeed. sml-insert-form is extensible so all you need to
do is create the macros yourself. Define a keybord macro
(C-x ( <something> C-x )) and give it a suitable name:
sml-addto-forms-alist prompts for a name, say NAME, and
binds the macro sml-form-NAME. Thereafter C-c RET
NAME will insert the macro at point, and C-u C-c RET NAME
will insert the macro after a newline-and-indent. If you want to
keep your macros from one editing session to the next, go to your
`.emacs' file and call insert-kbd-macro; you'll need
to add NAME to sml-forms-alist permanently yourself:
(defun my-sml-load-hook () "Global defaults for SML mode"
;; whatever else you do
(setq sml-forms-alist (cons '("NAME") sml-forms-alist)))
If you want to create templates like `case' that prompt for parameters
you'll have to do some Lisp programming. The tempo package looks
like a good stating point. You can always overwrite your own macros, but
the builtin forms for `let', etc., can't be overwritten.
I hate that indentation algorithm; can't I suppress it?
Ah, yes, a common complaint. It's actually very easy to use SML mode
without the troublesome sml-indent-line:
(defun my-sml-load-hook () "Global defaults for SML mode" ;; whatever else you do (fset 'sml-indent-line 'ignore))
though indent-relative-maybe may conceivable be more useful than
ignore.
The dedicated frame for ML is too huge; can it be made smaller?
Of course, you just have to modify the frame parameters. The variable
sml-display-frame-alist can be defined explicitly in your
`.emacs'; the default is a frame of 80 columns by 24 lines, and the
icon name will be the same as the ML interaction buffer's
name--something like *mosml*. I like a small, tidy font for this
frame so I
(setq sml-display-frame-alist
(cons '(font . "7x14") sml-display-frame-alist))
in my inferior-sml-load-hook. If you want fewer lines, try:
(setcdr (assoc 'height sml-display-frame-alist) 15)
or something.
Can SML mode handle more than one compiler running at once?
The question is whether you can! See the sml-buffer variable's
on-line help (C-h v sml-buffer). Note that the SML mode
compiler variables (see section Process defaults) are all buffer-local, so
you can even switch between different ML compilers, not just different
invocations of the same one. Well, you can.
What needs to be done to support other ML compilers?
Not that much really, at least not to create minimal support. The
interface between SML mode and the compiler is determined by the
variables
sml-use-command,
sml-cd-command,
sml-prompt-regexp
(which are easy to get right), and
sml-error-regexp, and
sml-error-parser (which are more tricky).
The general template to follow in setting this up
is in the files `sml-{poly-ml,mosml}.el'.
These rules will not change, I hope:
sml-next-error uses sml-error-regexp to locate the start
of the next error report in the ML interaction buffer (P)
sml-next-error calls sml-error-parser, passing P, and
expects up to five return values in this order:
eval'd
at (L,C) in F (EOE)
sml-error-parser can assume that P is the start of the next
error message that the user is interested in--since she defines this
point by defining sml-error-regexp.
sml-error-parser returns is a list. In the event of problems,
I foresee the following needs:
("std_in" L C)
(nil L C)
(F nil C)
(F L 1)
There's no need to return anything else. However, if you do want the
errorful text in F highlighted you should return a simple Lisp
expression in the fourth argument that'll compute the region. EOE
will be called with point at character (L,C) in F, and
should move point to the end of the errorful text. In fact, EOE
can actually do anything you wish, but in the simplest cases it'll just
(forward-char 45), or
(progn (forward-line 4) (forward-char 37))
etc.. If it does more, make sure it leaves point at the end of the
region in F---use save-excursion if switching buffers.
MSG, if returned, will be echoed in the minibuffer.
SML Mode was written originally by Lars Bo Nielsen for Emacs 18.5n;
later hacked for comint by Olin Shivers (who called it ml-mode);
much later hacked by myself because it didn't seem to work... Fritz
Knabe brilliantly posted the hilit19 and font-lock
functions on the net. Lars probably would recognise much of what
remains, yet now there're menus, syntax highlighting, support for
various ML compilers, Texinfo (hey!), and more than a little hope it'll
work with a variety of Emacs 19s. But there are still things to do. Lars
wrote:
The indentation algorithm still can be fooled. I don't know if it will ever be 100% right, as this means it will have to actually parse all of the buffer up to the actual line [...].
This is still the main cause of grief; SML's syntax is a nightmare for Emacs modes, and of course opinions vary about proper indentation. But there may be something we can do...
cf. commentary in the site initialisation file `sml-site.el'.
XEmacs 19.11 users are warned that
changing the default sml-temp-threshold may well cause XEmacs to
hang; they seem to have fixed the problem in 19.12 and above.
Does sml-error-parser return these
nice things? The answer is complicated! See section Advanced Topics, and the
docstring C-h v sml-error-parser.
ftp://sunsite.auc.dk/packages/auctex/
ftp://archive.cis.ohio-state.edu/pub/gnu/emacs/elisp-archive/misc/
This document was generated on 23 June 1997 using the texi2html translator version 1.51 (with some help by MJM)