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
22 {symbol = ' ', repr = ' '},
23 {symbol = '_', repr = '\\_'},
24 {symbol = '->>', repr = '\\twoheadrightarrow'},
25 {symbol = '->', repr = '→'},
26 {symbol = '=>', repr = '⇒'},
27 -- The default * sits very high above the baseline, \ast (u+2217) looks
29 {symbol = '*', repr = '\\ast'},
30 {symbol = '~', repr = '\\HDLine[width=.20 * \\the\\textwidth]'},
31 {symbol = '|', repr = '\\char' .. utf.byte('|')},
32 -- Use ▶ and ∧ from our roman font, since Iwona doesn't have the glyph
33 {symbol = '▶', repr = '{\\rm{}▶}'},
34 --{symbol = '∧', repr = '{$∧$}'},
35 {symbol = '∧', repr = '{\\rm{}∧}'},
38 -- Keywords that should be bold
46 ['DEFAULT'] = {small = true},
52 -- Store the last line for each indent level
53 local indentlines = {}
55 function array_concat(a1, a2)
57 for k,v in ipairs(a2) do
64 -- See if str starts with a symbol, and return the remaining string and that
65 -- symbol. If no symbol from the table is matched, just returns the first
66 -- character. We can do a lookup directly, since symbols can be different in
67 -- length, so we just loop over all symbols, trying them in turn.
68 local function take_symbol(str)
69 for i,props in ipairs(symbols) do
70 -- Try to remove symbol from the start of str
71 symbol, newstr = utf.match(str, "^(" .. props.symbol .. ")(.*)")
73 -- Return this tokens repr, or just the token if it has no
75 res = props.repr or symbol
76 -- Enclose the token in {\style .. }
78 res = "{\\" .. props.style .. " " .. res .. "}"
83 -- No symbol found, just return the first character
84 return utf.match(str, "^(.)(.*)")
87 -- Take a single word from str, if posible. Returns the rest of the string and
89 local function take_word(str)
90 -- A word must always start with a-z (in particular, λ is not a valid
92 res, newstr = utf.match(str, "^([a-zA-Z][%a%d%+%-%,_]+)(.*)")
93 return res, newstr or str
96 -- Tries to match each of the patterns and returns the captures of the first
97 -- matching pattern (up to 5 captures are supported). Returns nil when nothing
99 local function match_mul(str, patterns)
100 for i, pat in ipairs(patterns) do
101 a, b, c, d, e = utf.match(str, pat)
109 -- Find any subscripts in the given word and typeset them
110 local function do_subscripts(word)
111 base, sub = match_mul(res, submatches)
113 word = base .. "\\low{" .. sub .. "}"
114 -- After a word has been used as a base, allow subscripts
115 -- without _, even for non-numbers.
116 if not bases[base] then
117 -- Register that we've added this base
119 -- Add a patterns for this base. First, the base with a single
120 -- letter or number subscript.
121 submatches[#submatches+1] = "^(" .. base .. ")([%a%d])$"
122 -- Seconde, the base with a longer prefix that includes at least
123 -- one of +-, (to catch things like ri+1, but not return).
124 submatches[#submatches+1] = "^(" .. base .. ")([%a%d]*[%-%+,]+[%a%d%-%+,]*)$"
130 -- Do proper aligning for subsequent lines. For example, in
133 -- We replace the spaces in the second line with a skip with the same with as
134 -- "foo ", to align the | with the =.
135 -- For this, we keep a table "indentlines", which contains all previous lines
136 -- with smaller indent levels that are still "in scope" (e.g., have not yet
137 -- been followed by a line with a smaller indent level). For example:
143 -- After the last line, the table will contain:
144 -- { 0 = "line1", 2 = " line4", 4 = " line5"}
145 -- In other words, line3 is no longer in scope since it is "hidden" by
146 -- line4, and line is no longer in scope since it is replaced by line4.
147 local function do_indent(line)
148 newind, rest = utf.match(line, '^(%s*)(.*)')
150 -- Loop all the previous lines
151 for indent, unused in pairs(indentlines) do
152 if indent > #newind then
153 -- Remove any lines with a larger indent
154 indentlines[indent] = nil
155 elseif indent < #newind and indent > prev then
156 -- Find the last line (e.g, with the highest indent) with an
157 -- indent smaller than the new indent. This is the line from which
158 -- we need to copy the indent.
163 -- Always store this line, possibly overwriting a previous line with the
165 indentlines[#newind] = line
168 -- If there is a previous line with a smaller indent, make sure we
169 -- align with it. We do this by taking a prefix from that previous
170 -- line just as long as our indent. This gives us a bunch of
171 -- whitespace, with a few non-whitespace characters. We find out the
172 -- width of this prefix, and put whitespace just as wide as that
173 -- prefix before the current line, instead of the whitespace
174 -- characters that were there.
175 -- Doing this is slightly risky, since the prefix might contain
176 -- unfinished markup (e.g., \foo{bar without the closing }). We might
177 -- need to solve this later.
178 copyind = utf.sub(indentlines[prev], 1, #newind)
179 setwidth = "\\setwidthof{" .. copyind .. "}\\to\\pretlamalignwidth"
180 hskip = "\\hskip\\pretlamalignwidth"
181 return "{" .. setwidth .. hskip .. "}" .. rest
183 -- No previous line? Just return the unmodified line then
188 -- Mark the begin of a block of lambda formatted buffers or expressions. This
189 -- means that, until you call end_of_block again, the subscript bases are
190 -- shared. For example, if you have \lam{y1} some text \lam{yn} within a
191 -- single block, the yn will properly get subscripted. Be sure to call
192 -- end_of_block again!
194 -- Blocks can be partially nested, meaning that the block
195 -- won't be closed until end_of_block was called exactly as often as
196 -- begin_of_block. However, subscripts from the inner block can still
197 -- influence subscripts in the outer block.
198 function vis.begin_of_block()
199 vis.begin_of_display()
200 in_block = in_block + 1
203 -- Ends the current block
204 function vis.end_of_block()
205 in_block = in_block - 1
208 function vis.begin_of_display()
209 if in_block == 0 then
210 -- Initially allow subscripts using _ or just appending a number (later,
211 -- we will add extra patterns here.
212 submatches = {"^(%a*)_([%a%d,]+)$", "^(%a+)(%d[%d,]+)$"}
213 -- This stores all the bases we've encountered so far (to prevent
214 -- duplicates). For each of them there will be a pattern in submatches
222 -- Make things work for inline typeing (e.g., \type{}) as well.
223 vis.begin_of_inline = vis.begin_of_display
224 vis.end_of_inline = vis.end_of_display
226 function vis.flush_line(str,nested)
227 buffers.flush_result(vis.do_line(str, false), nested)
230 function vis.do_line(str, no_indent)
232 if not no_indent then
233 -- Allow ignore of the indentation stuff when we're calling ourselves
234 -- for a partial line.
240 local text, rest = utf.match(str, "^%-%-(.-)%-%-(.*)")
242 table.insert(result, '\\strikethrough{')
243 -- Recursively call ourselves to handle spaces gracefully.
244 result = array_concat(result, vis.do_line(text, true))
245 table.insert(result, '}')
246 -- Eat the processed characters
248 elseif utf.match(str, "^%-%-") then
249 table.insert(result, '{\\italic{--')
250 -- Recursively call ourselves to handle spaces gracefully.
251 result = array_concat(result, vis.do_line(utf.sub(str, 3), true))
252 table.insert(result, '}}')
253 -- Done with this line
256 -- See if the next token is a word
257 word, str = take_word(str)
259 if keywords[res] then
260 -- Make all keywords bold
261 word = "{\\bold " .. word .. "}"
262 if keywords[res].small then
263 word = "\\small" .. word -- Curlies were added above
266 -- Process any subscripts in the word
267 word = do_subscripts(word)
269 table.insert(result, word)
271 -- The next token is not a word, it must be a symbol
272 symbol, str = take_symbol(str)
273 table.insert(result, symbol)
281 -- vim: set sw=4 sts=4 expandtab ai: