1 -- filename : type-lam.lua
2 -- comment : Pretty printing of (extended) lambda calculus
3 -- author : Matthijs Kooijman, Universiteit Twente, NL
4 -- copyright: Matthijs Kooijman
7 local utf = unicode.utf8
9 local vis = buffers.newvisualizer("lam")
18 -- Symbols that should have a different representation
20 -- Note, the space we replace with is a Unicode non-breaking space
23 ['_'] = {repr = '\\_'},
24 ['->'] = {repr = '→'},
25 ['=>'] = {repr = '⇒'},
26 -- The default * sits very high above the baseline, \ast (u+2217) looks
28 ['*'] = {repr = '\\ast'},
29 ['~'] = {repr = '\\HDLine[width=.20 * \\the\\textwidth]'},
30 ['|'] = {repr = '\\char' .. utf.byte('|')},
31 ['$'] = {repr = '\\char' .. utf.byte('$')},
34 -- Keywords that should be bold
42 ['DEFAULT'] = {small = true},
48 -- Store the last line for each indent level
49 local indentlines = {}
51 -- See if str starts with a symbol, and return the remaining string and that
52 -- symbol. If no symbol from the table is matched, just returns the first
53 -- character. We can do a lookup directly, since symbols can be different in
54 -- length, so we just loop over all symbols, trying them in turn.
55 local function take_symbol(str)
56 for symbol,props in pairs(symbols) do
57 -- Try to remove symbol from the start of str
58 symbol, newstr = utf.match(str, "^(" .. symbol .. ")(.*)")
60 -- Return this tokens repr, or just the token if it has no
62 res = props.repr or symbol
63 -- Enclose the token in {\style .. }
65 res = "{\\" .. props.style .. " " .. res .. "}"
70 -- No symbol found, just return the first character
71 return utf.match(str, "^(.)(.*)")
74 -- Take a single word from str, if posible. Returns the rest of the string and
76 local function take_word(str)
77 -- A word must always start with a-z (in particular, λ is not a valid
79 res, newstr = utf.match(str, "^([a-zA-Z][%a%d%+%-%,_]+)(.*)")
80 return res, newstr or str
83 -- Tries to match each of the patterns and returns the captures of the first
84 -- matching pattern (up to 5 captures are supported). Returns nil when nothing
86 local function match_mul(str, patterns)
87 for i, pat in ipairs(patterns) do
88 a, b, c, d, e = utf.match(str, pat)
96 -- Find any subscripts in the given word and typeset them
97 local function do_subscripts(word)
98 base, sub = match_mul(res, submatches)
100 word = base .. "\\low{" .. sub .. "}"
101 -- After a word has been used as a base, allow subscripts
102 -- without _, even for non-numbers.
103 if not bases[base] then
104 -- Register that we've added this base
106 -- Add a patterns for this base. First, the base with a single
107 -- letter or number subscript.
108 submatches[#submatches+1] = "^(" .. base .. ")([%a%d])$"
109 -- Seconde, the base with a longer prefix that includes at least
110 -- one of +-, (to catch things like ri+1, but not return).
111 submatches[#submatches+1] = "^(" .. base .. ")([%a%d]*[%-%+,]+[%a%d%-%+,]*)$"
117 -- Do proper aligning for subsequent lines. For example, in
120 -- We replace the spaces in the second line with a skip with the same with as
121 -- "foo ", to align the | with the =.
122 -- For this, we keep a table "indentlines", which contains all previous lines
123 -- with smaller indent levels that are still "in scope" (e.g., have not yet
124 -- been followed by a line with a smaller indent level). For example:
130 -- After the last line, the table will contain:
131 -- { 0 = "line1", 2 = " line4", 4 = " line5"}
132 -- In other words, line3 is no longer in scope since it is "hidden" by
133 -- line4, and line is no longer in scope since it is replaced by line4.
134 local function do_indent(line)
135 newind, rest = utf.match(line, '^(%s*)(.*)')
137 -- Loop all the previous lines
138 for indent, unused in pairs(indentlines) do
139 if indent > #newind then
140 -- Remove any lines with a larger indent
141 indentlines[indent] = nil
142 elseif indent < #newind and indent > prev then
143 -- Find the last line (e.g, with the highest indent) with an
144 -- indent smaller than the new indent. This is the line from which
145 -- we need to copy the indent.
150 -- Always store this line, possibly overwriting a previous line with the
152 indentlines[#newind] = line
155 -- If there is a previous line with a smaller indent, make sure we
156 -- align with it. We do this by taking a prefix from that previous
157 -- line just as long as our indent. This gives us a bunch of
158 -- whitespace, with a few non-whitespace characters. We find out the
159 -- width of this prefix, and put whitespace just as wide as that
160 -- prefix before the current line, instead of the whitespace
161 -- characters that were there.
162 -- Doing this is slightly risky, since the prefix might contain
163 -- unfinished markup (e.g., \foo{bar without the closing }). We might
164 -- need to solve this later.
165 copyind = utf.sub(indentlines[prev], 1, #newind)
166 setwidth = "\\setwidthof{" .. copyind .. "}\\to\\pretlamalignwidth"
167 hskip = "\\hskip\\pretlamalignwidth"
168 return "{" .. setwidth .. hskip .. "}" .. rest
170 -- No previous line? Just return the unmodified line then
175 -- Mark the begin of a block of lambda formatted buffers or expressions. This
176 -- means that, until you call end_of_block again, the subscript bases are
177 -- shared. For example, if you have \lam{y1} some text \lam{yn} within a
178 -- single block, the yn will properly get subscripted. Be sure to call
179 -- end_of_block again!
181 -- Blocks can be partially nested, meaning that the block
182 -- won't be closed until end_of_block was called exactly as often as
183 -- begin_of_block. However, subscripts from the inner block can still
184 -- influence subscripts in the outer block.
185 function vis.begin_of_block()
186 vis.begin_of_display()
187 in_block = in_block + 1
190 -- Ends the current block
191 function vis.end_of_block()
192 in_block = in_block - 1
195 function vis.begin_of_display()
196 if in_block == 0 then
197 -- Initially allow subscripts using _ or just appending a number (later,
198 -- we will add extra patterns here.
199 submatches = {"^(%a*)_([%a%d,]+)$", "^(%a+)([%d,]+)$"}
200 -- This stores all the bases we've encountered so far (to prevent
201 -- duplicates). For each of them there will be a pattern in submatches
209 -- Make things work for inline typeing (e.g., \type{}) as well.
210 vis.begin_of_inline = vis.begin_of_display
211 vis.end_of_inline = vis.end_of_display
213 function vis.flush_line(str,nested)
214 local result, state = { }, 0
215 local finish, change = buffers.finish_state, buffers.change_state
217 -- Set the colorscheme, which is used by finish_state and change_state
218 buffers.currentcolors = colors
222 -- See if the next token is a word
223 word, str = take_word(str)
225 if keywords[res] then
226 -- Make all keywords bold
227 word = "{\\bold " .. word .. "}"
228 if keywords[res].small then
229 word = "\\small" .. word -- Curlies were added above
232 -- Process any subscripts in the word
233 word = do_subscripts(word)
236 -- The next token is not a word, it must be a symbol
237 symbol, str = take_symbol(str)
240 -- Append the resulting token
241 result[#result+1] = word or symbol
244 state = finish(state, result)
245 buffers.flush_result(result,nested)
248 -- vim: set sw=4 sts=4 expandtab ai: