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 ▶ from our roman font, since Iwona doesn't have the glyph
33 {symbol = '▶', repr = '{\\rm{}▶}'},
36 -- Keywords that should be bold
44 ['DEFAULT'] = {small = true},
50 -- Store the last line for each indent level
51 local indentlines = {}
53 function array_concat(a1, a2)
55 for k,v in ipairs(a2) do
62 -- See if str starts with a symbol, and return the remaining string and that
63 -- symbol. If no symbol from the table is matched, just returns the first
64 -- character. We can do a lookup directly, since symbols can be different in
65 -- length, so we just loop over all symbols, trying them in turn.
66 local function take_symbol(str)
67 for i,props in ipairs(symbols) do
68 -- Try to remove symbol from the start of str
69 symbol, newstr = utf.match(str, "^(" .. props.symbol .. ")(.*)")
71 -- Return this tokens repr, or just the token if it has no
73 res = props.repr or symbol
74 -- Enclose the token in {\style .. }
76 res = "{\\" .. props.style .. " " .. res .. "}"
81 -- No symbol found, just return the first character
82 return utf.match(str, "^(.)(.*)")
85 -- Take a single word from str, if posible. Returns the rest of the string and
87 local function take_word(str)
88 -- A word must always start with a-z (in particular, λ is not a valid
90 res, newstr = utf.match(str, "^([a-zA-Z][%a%d%+%-%,_]+)(.*)")
91 return res, newstr or str
94 -- Tries to match each of the patterns and returns the captures of the first
95 -- matching pattern (up to 5 captures are supported). Returns nil when nothing
97 local function match_mul(str, patterns)
98 for i, pat in ipairs(patterns) do
99 a, b, c, d, e = utf.match(str, pat)
107 -- Find any subscripts in the given word and typeset them
108 local function do_subscripts(word)
109 base, sub = match_mul(res, submatches)
111 word = base .. "\\low{" .. sub .. "}"
112 -- After a word has been used as a base, allow subscripts
113 -- without _, even for non-numbers.
114 if not bases[base] then
115 -- Register that we've added this base
117 -- Add a patterns for this base. First, the base with a single
118 -- letter or number subscript.
119 submatches[#submatches+1] = "^(" .. base .. ")([%a%d])$"
120 -- Seconde, the base with a longer prefix that includes at least
121 -- one of +-, (to catch things like ri+1, but not return).
122 submatches[#submatches+1] = "^(" .. base .. ")([%a%d]*[%-%+,]+[%a%d%-%+,]*)$"
128 -- Do proper aligning for subsequent lines. For example, in
131 -- We replace the spaces in the second line with a skip with the same with as
132 -- "foo ", to align the | with the =.
133 -- For this, we keep a table "indentlines", which contains all previous lines
134 -- with smaller indent levels that are still "in scope" (e.g., have not yet
135 -- been followed by a line with a smaller indent level). For example:
141 -- After the last line, the table will contain:
142 -- { 0 = "line1", 2 = " line4", 4 = " line5"}
143 -- In other words, line3 is no longer in scope since it is "hidden" by
144 -- line4, and line is no longer in scope since it is replaced by line4.
145 local function do_indent(line)
146 newind, rest = utf.match(line, '^(%s*)(.*)')
148 -- Loop all the previous lines
149 for indent, unused in pairs(indentlines) do
150 if indent > #newind then
151 -- Remove any lines with a larger indent
152 indentlines[indent] = nil
153 elseif indent < #newind and indent > prev then
154 -- Find the last line (e.g, with the highest indent) with an
155 -- indent smaller than the new indent. This is the line from which
156 -- we need to copy the indent.
161 -- Always store this line, possibly overwriting a previous line with the
163 indentlines[#newind] = line
166 -- If there is a previous line with a smaller indent, make sure we
167 -- align with it. We do this by taking a prefix from that previous
168 -- line just as long as our indent. This gives us a bunch of
169 -- whitespace, with a few non-whitespace characters. We find out the
170 -- width of this prefix, and put whitespace just as wide as that
171 -- prefix before the current line, instead of the whitespace
172 -- characters that were there.
173 -- Doing this is slightly risky, since the prefix might contain
174 -- unfinished markup (e.g., \foo{bar without the closing }). We might
175 -- need to solve this later.
176 copyind = utf.sub(indentlines[prev], 1, #newind)
177 setwidth = "\\setwidthof{" .. copyind .. "}\\to\\pretlamalignwidth"
178 hskip = "\\hskip\\pretlamalignwidth"
179 return "{" .. setwidth .. hskip .. "}" .. rest
181 -- No previous line? Just return the unmodified line then
186 -- Mark the begin of a block of lambda formatted buffers or expressions. This
187 -- means that, until you call end_of_block again, the subscript bases are
188 -- shared. For example, if you have \lam{y1} some text \lam{yn} within a
189 -- single block, the yn will properly get subscripted. Be sure to call
190 -- end_of_block again!
192 -- Blocks can be partially nested, meaning that the block
193 -- won't be closed until end_of_block was called exactly as often as
194 -- begin_of_block. However, subscripts from the inner block can still
195 -- influence subscripts in the outer block.
196 function vis.begin_of_block()
197 vis.begin_of_display()
198 in_block = in_block + 1
201 -- Ends the current block
202 function vis.end_of_block()
203 in_block = in_block - 1
206 function vis.begin_of_display()
207 if in_block == 0 then
208 -- Initially allow subscripts using _ or just appending a number (later,
209 -- we will add extra patterns here.
210 submatches = {"^(%a*)_([%a%d,]+)$", "^(%a+)(%d[%d,]+)$"}
211 -- This stores all the bases we've encountered so far (to prevent
212 -- duplicates). For each of them there will be a pattern in submatches
220 -- Make things work for inline typeing (e.g., \type{}) as well.
221 vis.begin_of_inline = vis.begin_of_display
222 vis.end_of_inline = vis.end_of_display
224 function vis.flush_line(str,nested)
225 buffers.flush_result(vis.do_line(str, false), nested)
228 function vis.do_line(str, no_indent)
230 if not no_indent then
231 -- Allow ignore of the indentation stuff when we're calling ourselves
232 -- for a partial line.
238 local text, rest = utf.match(str, "^%-%-(.-)%-%-(.*)")
240 table.insert(result, '\\strikethrough{')
241 -- Recursively call ourselves to handle spaces gracefully.
242 result = array_concat(result, vis.do_line(text, true))
243 table.insert(result, '}')
244 -- Eat the processed characters
246 elseif utf.match(str, "^%-%-") then
247 table.insert(result, '{\\italic{--')
248 -- Recursively call ourselves to handle spaces gracefully.
249 result = array_concat(result, vis.do_line(utf.sub(str, 3), true))
250 table.insert(result, '}}')
251 -- Done with this line
254 -- See if the next token is a word
255 word, str = take_word(str)
257 if keywords[res] then
258 -- Make all keywords bold
259 word = "{\\bold " .. word .. "}"
260 if keywords[res].small then
261 word = "\\small" .. word -- Curlies were added above
264 -- Process any subscripts in the word
265 word = do_subscripts(word)
267 table.insert(result, word)
269 -- The next token is not a word, it must be a symbol
270 symbol, str = take_symbol(str)
271 table.insert(result, symbol)
279 -- vim: set sw=4 sts=4 expandtab ai: