Improve subscript handling in pret-lam.

The handling is now a bit more strict when automatically detecting
multiple character subscripts, to prevent things like "return" from being
turned into "r_eturn".

diff --git a/pret-lam.lua b/pret-lam.lua
index aff6988d8af774ef8f1e1021f954ed3f34e786d9..ea07811759ae40920b5ef4d6a4f8e174a3e3fa1f 100644 (file)
--- a/pret-lam.lua
+++ b/pret-lam.lua
@@ -63,7 +63,7 @@ end
 local function take_word(str)
         -- A word must always start with a-z (in particular, λ is not a valid
         -- start of a word).
-        res, newstr = utf.match(str, "^([a-zA-Z][%a%d_]+)(.*)")
+        res, newstr = utf.match(str, "^([a-zA-Z][%a%d%+%-%,_]+)(.*)")
         return res, newstr or str
 end
 
@@ -90,8 +90,12 @@ local function do_subscripts(word)
         if not bases[base] then
             -- Register that we've added this base
             bases[base] = true
-            -- Add a pattern for this base
-            submatches[#submatches+1] = "^(" .. base .. ")([%a%d,]+)$"
+            -- Add a patterns for this base. First, the base with a single
+            -- letter or number subscript.
+            submatches[#submatches+1] = "^(" .. base .. ")([%a%d])$"
+            -- Seconde, the base with a longer prefix that includes at least
+            -- one of +-, (to catch things like ri+1, but not return).
+            submatches[#submatches+1] = "^(" .. base .. ")([%a%d]*[%-%+%,]+[%a%d%-%+%,]*)$"
         end
     end
     return word
@@ -100,7 +104,7 @@ end
 function vis.begin_of_display()
     -- Initially allow subscripts using _ or just appending a number (later,
     -- we will add extra patterns here.
-    submatches = {"^(.*)_([%a%d,]+)$", "^(.*[^%d])(%d+)$"}
+    submatches = {"^(%a*)_([%a%d,]+)$", "^(%a+)(%d+)$"}
     -- This stores all the bases we've encountered so far (to prevent
     -- duplicates). For each of them there will be a pattern in submatches
     -- above.