textboxwidget: even better text wrapping

util.isSplitable() accepts now also the previous char to help
decide if a space can be used to split a line.
TextBoxWidget:_splitCharWidthList() : simplified logic

frontend/ui/widget/textboxwidget.lua

@@ -118,31 +118,32 @@ function TextBoxWidget:_splitCharWidthList()
             cur_line_text = table.concat(self.charlist, "", offset, idx - 1)
         else
             -- Backtrack the string until the length fit into one line.
+            -- We'll give next and prev chars to isSplitable() for a wiser decision
             local c = self.char_width_list[idx].char
-            -- We give next char to isSplitable() for a wiser decision
+            local next_c = idx+1 <= size and self.char_width_list[idx+1].char or false
-            local next_c = idx+1 <= size and self.char_width_list[idx+1].char or nil
+            local prev_c = idx-1 >= 1 and self.char_width_list[idx-1].char or false
-            if util.isSplitable(c, next_c) then
+            local adjusted_idx = idx
+            local adjusted_width = cur_line_width
+            while adjusted_idx > offset and not util.isSplitable(c, next_c, prev_c) do
+                adjusted_width = adjusted_width - self.char_width_list[adjusted_idx].width
+                adjusted_idx = adjusted_idx - 1
+                next_c = c
+                c = prev_c
+                prev_c = adjusted_idx-1 >= 1 and self.char_width_list[adjusted_idx-1].char or false
+            end
+            if adjusted_idx == offset or adjusted_idx == idx then
+                -- either a very long english word ocuppying more than one line,
+                -- or the excessive char is itself splitable:
+                -- we let that excessive char for next line
                 cur_line_text = table.concat(self.charlist, "", offset, idx - 1)
                 cur_line_width = cur_line_width - self.char_width_list[idx].width
             else
-                local adjusted_idx = idx
+                -- we backtracked and we're below max width, we can let the
-                local adjusted_width = cur_line_width
+                -- splitable char on this line
-                repeat
+                cur_line_text = table.concat(self.charlist, "", offset, adjusted_idx)
-                    adjusted_width = adjusted_width - self.char_width_list[adjusted_idx].width
+                cur_line_width = adjusted_width
-                    if adjusted_idx == 1 then break end
+                idx = adjusted_idx + 1
-                    adjusted_idx = adjusted_idx - 1
+            end
-                    next_c = c
-                    c = self.char_width_list[adjusted_idx].char
-                until adjusted_idx == offset or util.isSplitable(c, next_c)
-                if adjusted_idx == offset then -- a very long english word ocuppying more than one line
-                    cur_line_text = table.concat(self.charlist, "", offset, idx - 1)
-                    cur_line_width = cur_line_width - self.char_width_list[idx].width
-                else
-                    cur_line_text = table.concat(self.charlist, "", offset, adjusted_idx)
-                    cur_line_width = adjusted_width
-                    idx = adjusted_idx + 1
-                end
-            end -- endif util.isSplitable(c)
         end -- endif cur_line_width > self.width
         if cur_line_width < 0 then break end
         self.vertical_string_list[ln] = {

frontend/util.lua

@@ -150,22 +150,24 @@ end
 -- specific punctuation : e.g. "word :" or "word )"
 -- (In french, there is a space before a colon, and it better
 -- not be wrapped there.)
--- Includes U+00BB >> (right double angle quotation mark) and
+local non_splitable_space_tailers = ":;,.!?)]}$%=-+*/|<>»”"
--- U+201D '' (right double quotation mark)
+-- Same if a space has some specific other punctuation before it
-local non_splitable_space_tailers = ":;,.!?)]}$%-=/<>»”"
+local non_splitable_space_leaders = "([{$=-+*/|<>«“"
 
 -- Test whether a string could be separated by this char for multi-line rendering
--- Optional next char may be provided to help make the decision
+-- Optional next or prev chars may be provided to help make the decision
-function util.isSplitable(c, next_c)
+function util.isSplitable(c, next_c, prev_c)
     if util.isCJKChar(c) then
         -- a CJKChar is a word in itself, and so is splitable
         return true
     elseif c == " " then
         -- we only split on a space (so punctuation sticks to prev word)
-        -- if next_c is provided, we can make a better decision
+        -- if next_c or prev_c is provided, we can make a better decision
         if next_c and non_splitable_space_tailers:find(next_c, 1, true) then
-            -- this space is followed by some punctuation that is better
+            -- this space is followed by some punctuation that is better kept with us
-            -- kept with us along previous word
+            return false
+        elseif prev_c and non_splitable_space_leaders:find(prev_c, 1, true) then
+            -- this space is lead by some punctuation that is better kept with us
             return false
         else
             -- we can split on this space

spec/unit/util_spec.lua

@@ -157,5 +157,38 @@ describe("util module", function()
         })
     end)
 
+    it("should split text to line with next_c and prev_c - unicode", function()
+        local text = "Ce test : 1) est « très simple » ; 2 ) simple comme ( 2/2 ) > 50 % ? ok."
+        local word = ""
+        local table_of_words = {}
+        local c
+        local table_chars = util.splitToChars(text)
+        for i = 1, #table_chars  do
+            c = table_chars[i]
+            next_c = i < #table_chars and table_chars[i+1] or nil
+            prev_c = i > 1 and table_chars[i-1] or nil
+            word = word .. c
+            if util.isSplitable(c, next_c, prev_c) then
+                table.insert(table_of_words, word)
+                word = ""
+            end
+            if i == #table_chars then table.insert(table_of_words, word) end
+        end
+        assert.are_same(table_of_words, {
+            "Ce ",
+            "test : ",
+            "1) ",
+            "est ",
+            "« très ",
+            "simple » ; ",
+            "2 ) ",
+            "simple ",
+            "comme ",
+            "( 2/2 ) > 50 % ? ",
+            "ok."
+        })
+    end)
+
+
 
 end)