X-Git-Url: https://git.stderr.nl/gitweb?p=matthijs%2Fmaster-project%2Freport.git;a=blobdiff_plain;f=pret-lam.lua;h=68db5772120fcf8716ada9e40b3f612a43986747;hp=c25daefb697c3cd08c7a15115c935a62a948cf42;hb=651af924b4870140446c9f3551b3aea1095d8987;hpb=72702b3b4183aa6cf75fe01ae5a07aaf85790801

diff --git a/pret-lam.lua b/pret-lam.lua
index c25daef..68db577 100644
--- a/pret-lam.lua
+++ b/pret-lam.lua
@@ -6,57 +6,234 @@
 
 local utf = unicode.utf8
 
-if not buffers                 then buffers                 = { } end
-if not buffers.visualizers     then buffers.visualizers     = { } end
-if not buffers.visualizers.lam then buffers.visualizers.lam = { } end
+local vis = buffers.newvisualizer("lam")
 
-buffers.visualizers.lam.colors = {
+local colors = {
     "prettytwo",
     "prettyone",
     "prettythree",
     "prettyfour"
 }
 
-buffers.visualizers.lam.tokens = {
+-- Symbols that should have a different representation
+local symbols = {
     [' '] = {repr = '\\obs '},
     ['_'] = {repr = '\\_'},
     ['->'] = {repr = '\\rightarrow'},
-    ['case'] = {style = 'bold'},
-    ['of'] = {style = 'bold'},
-    ['let'] = {style = 'bold'},
-    ['in'] = {style = 'bold'},
+    -- The default * sits very high above the baseline, \ast (u+2217) looks
+    -- better.
+    ['*'] = {repr = '\\ast'},
+    ['~'] = {repr = '\\HDLine[width=.20 * \\the\\textwidth]'},
+    ['|'] = {repr = '\\char' .. utf.byte('|')},
 }
 
-function buffers.visualizers.lam.flush_line(str,nested)
+-- Keywords that should be bold
+local keywords = {
+    ['case'] = {},
+    ['of'] = {},
+    ['let'] = {},
+    ['letrec'] = {},
+    ['letnonrec'] = {},
+    ['in'] = {},
+    ['DEFAULT'] = {},
+}
+
+local in_block = 0
+local submatches = {}
+local bases = {}
+-- Store the last line for each indent level
+local indentlines = {}
+
+-- See if str starts with a symbol, and return the remaining string and that
+-- symbol. If no symbol from the table is matched, just returns the first
+-- character.  We can do a lookup directly, since symbols can be different in
+-- length, so we just loop over all symbols, trying them in turn.
+local function take_symbol(str)
+    for symbol,props in pairs(symbols) do
+        -- Try to remove symbol from the start of str 
+        symbol, newstr = utf.match(str, "^(" .. symbol .. ")(.*)")
+        if symbol then
+            -- Return this tokens repr, or just the token if it has no
+            -- repr.
+            res = props.repr or symbol
+            -- Enclose the token in {\style .. }
+            if props.style then
+                res = "{\\" .. props.style ..  " " .. res ..  "}"
+            end
+            return res, newstr
+        end
+    end
+    -- No symbol found, just return the first character
+    return utf.match(str, "^(.)(.*)")
+end
+
+-- Take a single word from str, if posible. Returns the rest of the string and
+-- the word taken.
+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%+%-%,_]+)(.*)")
+        return res, newstr or str
+end
+
+-- Tries to match each of the patterns and returns the captures of the first
+-- matching pattern (up to 5 captures are supported). Returns nil when nothing
+-- matches.
+local function match_mul(str, patterns)
+    for i, pat in ipairs(patterns) do
+        a, b, c, d, e = utf.match(str, pat)
+        if a then
+            return a, b, c, d, e
+        end
+    end
+    return nil
+end
+
+-- Find any subscripts in the given word and typeset them
+local function do_subscripts(word)
+    base, sub = match_mul(res, submatches)
+    if sub then
+        word = base .. "\\low{" .. sub .. "}"
+        -- After a word has been used as a base, allow subscripts
+        -- without _, even for non-numbers.
+        if not bases[base] then
+            -- Register that we've added this base
+            bases[base] = true
+            -- 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
+end
+
+-- Do proper aligning for subsequent lines. For example, in 
+--   foo = bar
+--       | baz
+-- We replace the spaces in the second line with a skip with the same with as
+-- "foo ", to align the | with the =.
+-- For this, we keep a table "indentlines", which contains all previous lines
+-- with smaller indent levels that are still "in scope" (e.g., have not yet
+-- been followed by a line with a smaller indent level). For example:
+--   line1
+--     line2
+--       line3
+--     line4
+--       line5
+-- After the last line, the table will contain:
+--   { 0 = "line1", 2 = "  line4", 4 = "    line5"}
+--   In other words, line3 is no longer in scope since it is "hidden" by
+--   line4, and line is no longer in scope since it is replaced by line4.
+local function do_indent(line)
+    newind, rest = utf.match(line, '^(%s*)(.*)')
+    prev = -1
+    -- Loop all the previous lines
+    for indent, unused in pairs(indentlines) do
+        if indent > #newind then
+            -- Remove any lines with a larger indent
+            indentlines[indent] = nil
+        elseif indent < #newind and indent > prev then
+            -- Find the last line (e.g, with the highest indent) with an
+            -- indent smaller than the new indent. This is the line from which
+            -- we need to copy the indent.
+            prev = indent
+        end
+    end
+    
+    -- Always store this line, possibly overwriting a previous line with the
+    -- same indent
+    indentlines[#newind] = line
+
+    if prev ~= -1 then
+        -- If there is a previous line with a smaller indent, make sure we
+        -- align with it. We do this by taking a prefix from that previous
+        -- line just as long as our indent. This gives us a bunch of
+        -- whitespace, with a few non-whitespace characters. We find out the
+        -- width of this prefix, and put whitespace just as wide as that
+        -- prefix before the current line, instead of the whitespace
+        -- characters that were there.
+        -- Doing this is slightly risky, since the prefix might contain
+        -- unfinished markup (e.g., \foo{bar without the closing }). We might
+        -- need to solve this later.
+        copyind = utf.sub(indentlines[prev], 1, #newind)
+        setwidth = "\\setwidthof{" .. copyind .. "}\\to\\pretlamalignwidth"
+        hskip = "\\hskip\\pretlamalignwidth"
+        return "{" .. setwidth .. hskip .. "}" .. rest
+    end
+        -- No previous line? Just return the unmodified line then
+        return line
+end
+
+
+-- Mark the begin of a block of lambda formatted buffers or expressions. This
+-- means that, until you call end_of_block again, the subscript bases are
+-- shared. For example, if you have \lam{y1} some text \lam{yn} within a
+-- single block, the yn will properly get subscripted. Be sure to call
+-- end_of_block again!
+--
+-- Blocks can be partially nested, meaning that the block
+-- won't be closed until end_of_block was called exactly as often as
+-- begin_of_block. However, subscripts from the inner block can still
+-- influence subscripts in the outer block.
+function vis.begin_of_block()
+    vis.begin_of_display()
+    in_block = in_block + 1
+end
+
+-- Ends the current block
+function vis.end_of_block()
+    in_block = in_block - 1
+end
+
+function vis.begin_of_display()
+    if in_block == 0 then
+        -- Initially allow subscripts using _ or just appending a number (later,
+        -- we will add extra patterns here.
+        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.
+        bases = {}
+    end
+    indentlines = {}
+end
+    
+
+-- Make things work for inline typeing (e.g., \type{}) as well.
+vis.begin_of_inline = vis.begin_of_display
+vis.end_of_inline = vis.end_of_display
+
+function vis.flush_line(str,nested)
     local result, state = { }, 0
     local finish, change = buffers.finish_state, buffers.change_state
+    str = do_indent(str)
     -- Set the colorscheme, which is used by finish_state and change_state
-    buffers.currentcolors = buffers.visualizers.lam.colors
-
+    buffers.currentcolors = colors
     while str ~= "" do
-        found = false
-        for tok,props in pairs(buffers.visualizers.lam.tokens) do
-            -- Try to remove tok from the start of str 
-            str, count = utf.gsub(str, "^" .. tok, "")
-            if count ~= 0 then
-                -- Replace the token with the repr, or just itself
-                result[#result+1] = props.repr or tok
-                -- Enclose the token in {\style .. }
-                if props.style then
-                    result[#result] = "{\\" .. props.style ..
-                                        " " .. result[#result] ..
-                                        "}"
-                end
-                -- We found a token, now start over
-                found = true
-                break
+        local found = false
+        local word, symbol
+        -- See if the next token is a word
+        word, str = take_word(str)
+        if word then
+            if keywords[res] then
+                -- Make all keywords bold
+                word = "{\\bold " .. word ..  "}"
+            else
+                -- Process any subscripts in the word
+                word = do_subscripts(word)
             end
+        else
+            -- The next token is not a word, it must be a symbol
+            symbol, str = take_symbol(str)
         end
-        if not found then
-            result[#result+1] = utf.sub(str, 1, 1)
-            str = utf.sub(str, 2)
-        end
+
+        -- Append the resulting token
+        result[#result+1] = word or symbol
     end
+    
     state = finish(state, result)
     buffers.flush_result(result,nested)
 end