-- filename : type-lam.lua -- comment : Pretty printing of (extended) lambda calculus -- author : Matthijs Kooijman, Universiteit Twente, NL -- copyright: Matthijs Kooijman -- license : None local utf = unicode.utf8 local vis = buffers.newvisualizer("lam") local colors = { "prettytwo", "prettyone", "prettythree", "prettyfour" } -- Symbols that should have a different representation local symbols = { [' '] = {repr = '\\obs '}, ['_'] = {repr = '\\_'}, ['->'] = {repr = '\\rightarrow'}, -- 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('|')}, } -- 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 = {} -- The amount of indent of the first line, which we will strip of all -- subsequent lines. This allows the entire block to be indented as normal in -- the tex source code. local first_indent -- 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*)(.*)') -- Store the first line's indent if not first_indent then first_indent = utf.len(newind) end -- Strip the indent of the first line from this line's indent. newind = utf.sub(newind, first_indent + 1) -- Rebuild line, so we can still use it below line = newind .. rest 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 = {} first_indent = nil 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 = colors while str ~= "" do 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 -- Append the resulting token result[#result+1] = word or symbol end state = finish(state, result) buffers.flush_result(result,nested) end -- vim: set sw=4 sts=4 expandtab ai: