Satish B. SettyArchiveAboutRSS Feed

Custom input method for Emacs

There are different input methods built into Emacs (M-x set-input-method) for different languages and scripts. The ones I commonly use are:

  1. latin-postfix
  2. devanagari-itrans
  3. kannada-itrans
  4. telugu-itrans

More information is available through M-x describe-input-method. However, there is no built-in facility to input IAST, which is probably the most common transliteration scheme for Sanskrit and other Indic languages.

So, this is a short tutorial on how to define your own custom input method, with IAST as example.


Quail is minor mode for inputting multilingual text. It provides useful functions for abstracting input methods. The idea is that transliteration rules are stored as a sort of key-value pairs (aka hash table). Quail will take care of converting this table into a lower level suitable for Emacs.

There are just two functions, quail-define-package and quail-define-rules, that we must implement. The latter must follow the former immediately. The parameters for transliteration are specified in the former where as the mapping table itself is specified in the latter.


This function takes three mandatory paramters and a dozen optional ones.

(quail-define-package NAME LANGUAGE TITLE 
        &optional GUIDANCE DOCSTRING ... SIMPLE)

The mandatory paramters NAME, LANGUAGE and TITLE are self-explanatory. GUIDANCE is a boolean. If it’s set to true, Emacs will show all possible completions for the keystrokes in the echo area, right while you’re typing. This is useful for resolving ambiguities or getting quick help. The last parameter SIMPLE is also a boolean. Setting it to true means we’re promising Emacs that we’ll not mess with key bindings like C-b and C-f which are used for navigation. We shall set all other params to nil.

Here’s how it looks now

 "iast-postfix" "UTF-8" "InR<" t
  "Input method for Indic transliteration with postfix modifiers.

     Long vowels are dealt with by doubling.

     |                  | postfix | examples             |
     | macron           |         | aa  -> ā    ee  -> ē |
     | diacritic below  | .       | d.  -> ḍ    rr. -> ṝ |
     | diacritic above  | '       | s'  -> ś    n'  -> ṅ |
     | tilde            | ~       | n~  -> ñ             |
  nil t nil nil nil nil nil nil nil nil t)

We’re saying that our new input method “iast-postfix” uses UTF–8 encoding. TITLE is InR< (Indic Roman, < means postfix) which will be shown in Emacs’ mode line for highlighting the current input method. Docstring follows next. The last param SIMPLE is set to true t. DOCSTRING is free text that defines the format of input method which we’ve designed.


This part is easy. Just put the mapping table! It’s very similar to m17n’s syntax (see for example sa-iast.mim), which is no coincidence because Kenichi HANDA maintains both of them.

 ;; long vowels
 ("aa" "ā")
 ("ii" "ī")
 ("uu" "ū")
 ("rr." "ṝ")
 ("ee" "ē")
 ("oo" "ō")

 ;; dot below
 ("r." "ṛ")
 ("l." "ḷ")
 ("m." "ṃ")
 ("h." "ḥ")
 ("t." "ṭ")
 ("d." "ḍ")
 ("n." "ṇ")
 ("s." "ṣ")
 ;; diacritic above
 ("n'" "ṅ")
 ("s'" "ś")
 ("n~" "ñ")

Obviously you can extend this table to have uppercase characters, etc. Note that the second elemen in each of the pairs above is a pre-composed Unicode codepoint. If it consists of multiple codepoints, you need to use square brackets:

("gy" ["jñ"])  ; as in, gyaana becomes jñāna

Loading the input method

Just save the above two functions in a file, say “indic-input.el”. Convert that file into an Emacs package by simply adding this as the last line of the file:

(provide 'indic-input)

You can then use the above package by putting these lines in your emacs init file:

(add-to-list 'load-path "/folder/where/this/file/exists/")
(require 'indic-input)

You can then switch to the input method in any buffer by the usual means M-x set-input-method and choosing iast-postfix.