For writing texts on spectroscopy I would like to use abbrevs for common notations for spectral lines. However, the method outlined in the documentation for defining abbrevs with C-x a l or C-x a g fails on me. For example, I type Ly M-x insert-char greek small letter alpha to get Lyα now I hit C-x a l and only α is displayed in the minibuffer. Why? Is the only solution for me to manually edit the abbrev-table?

1 Answer 1


Curious. add-abbrev scans backward by words using forward-word to find out what abbrev to define. I'm not entirely sure why, but even though both "y" and "α" are defined as word characters in the syntax table (at least in the modes I tried it in, ymmv), it still considers there to be a word boundary between them. You can see this if you skip backwards and forwards by words with M-b or M-f. This is implemented in the scan_words function in syntax.c, which is sufficiently obscure that I can't tell exactly what the cause of this is; it might even be a feature.

Anyway, the help for add-global-abbrev (and add-local-abbrev) says this:

The prefix argument specifies the number of words before point that form the
expansion; or zero means the region is the expansion.
A negative argument means to undefine the specified abbrev.

This means that you can specify a prefix argument to tell it how many words to use; C-u 2 C-x a g does the trick.


I figured it out. scan_words considers two word characters to have a word boundary between them if they are from different scripts. See word_boundary_p in category.c, and the variable char-script-table which maps characters to script names. I suppose that makes a reasonable rule to use.

I suppose you could advise add-abbrev so that it uses forward-word-strictly, which ignores the script and goes only by the syntax table. Or you could mess around with find-word-boundary-function-table, but that seems like a lot of work.

  • 2
    For this case you'd want to use word-combining-categories, eg (add-to-list 'word-combining-categories (cons ?l ?G)) to indicate that Latin letters followed by Greek letters don't form a word boundary
    – rpluim
    Commented Apr 3, 2020 at 9:16
  • Thanks for your answer @db48x , this helped. I am still confused, this is the second time I play with abbrev mode, and seems to be never doing what I like. Try entering b M-x insert-char less-than or equivalent to, which expands to b ≲, now C-x a l wants to define an abbrev for b ≲ and not just (I havent altered the word combining categories). So the edit to your answer seems not to be correct... Commented Apr 6, 2020 at 23:14
  • 1
    Unicode classifies that as a symbol rather than as a word-constituent. If you use backword-word (M-b), you'll see that it jumps over all punctuation to get to the start of the previous word. Since add-abbrev calls backward-word to find the start of the previous word, it does so as well.
    – db48x
    Commented Apr 7, 2020 at 2:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.