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 ['$'] = {repr = '\\char' .. utf.byte('$')},
32 -- Use ▶ from our roman font, since Iwona doesn't have the glyph
33 ['▶'] = {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 -- See if str starts with a symbol, and return the remaining string and that
54 -- symbol. If no symbol from the table is matched, just returns the first
55 -- character. We can do a lookup directly, since symbols can be different in
56 -- length, so we just loop over all symbols, trying them in turn.
57 local function take_symbol(str)
58 for symbol,props in pairs(symbols) do
59 -- Try to remove symbol from the start of str
60 symbol, newstr = utf.match(str, "^(" .. symbol .. ")(.*)")
62 -- Return this tokens repr, or just the token if it has no
64 res = props.repr or symbol
65 -- Enclose the token in {\style .. }
67 res = "{\\" .. props.style .. " " .. res .. "}"
72 -- No symbol found, just return the first character
73 return utf.match(str, "^(.)(.*)")
76 -- Take a single word from str, if posible. Returns the rest of the string and
78 local function take_word(str)
79 -- A word must always start with a-z (in particular, λ is not a valid
81 res, newstr = utf.match(str, "^([a-zA-Z][%a%d%+%-%,_]+)(.*)")
82 return res, newstr or str
85 -- Tries to match each of the patterns and returns the captures of the first
86 -- matching pattern (up to 5 captures are supported). Returns nil when nothing
88 local function match_mul(str, patterns)
89 for i, pat in ipairs(patterns) do
90 a, b, c, d, e = utf.match(str, pat)
98 -- Find any subscripts in the given word and typeset them
99 local function do_subscripts(word)
100 base, sub = match_mul(res, submatches)
102 word = base .. "\\low{" .. sub .. "}"
103 -- After a word has been used as a base, allow subscripts
104 -- without _, even for non-numbers.
105 if not bases[base] then
106 -- Register that we've added this base
108 -- Add a patterns for this base. First, the base with a single
109 -- letter or number subscript.
110 submatches[#submatches+1] = "^(" .. base .. ")([%a%d])$"
111 -- Seconde, the base with a longer prefix that includes at least
112 -- one of +-, (to catch things like ri+1, but not return).
113 submatches[#submatches+1] = "^(" .. base .. ")([%a%d]*[%-%+,]+[%a%d%-%+,]*)$"
119 -- Do proper aligning for subsequent lines. For example, in
122 -- We replace the spaces in the second line with a skip with the same with as
123 -- "foo ", to align the | with the =.
124 -- For this, we keep a table "indentlines", which contains all previous lines
125 -- with smaller indent levels that are still "in scope" (e.g., have not yet
126 -- been followed by a line with a smaller indent level). For example:
132 -- After the last line, the table will contain:
133 -- { 0 = "line1", 2 = " line4", 4 = " line5"}
134 -- In other words, line3 is no longer in scope since it is "hidden" by
135 -- line4, and line is no longer in scope since it is replaced by line4.
136 local function do_indent(line)
137 newind, rest = utf.match(line, '^(%s*)(.*)')
139 -- Loop all the previous lines
140 for indent, unused in pairs(indentlines) do
141 if indent > #newind then
142 -- Remove any lines with a larger indent
143 indentlines[indent] = nil
144 elseif indent < #newind and indent > prev then
145 -- Find the last line (e.g, with the highest indent) with an
146 -- indent smaller than the new indent. This is the line from which
147 -- we need to copy the indent.
152 -- Always store this line, possibly overwriting a previous line with the
154 indentlines[#newind] = line
157 -- If there is a previous line with a smaller indent, make sure we
158 -- align with it. We do this by taking a prefix from that previous
159 -- line just as long as our indent. This gives us a bunch of
160 -- whitespace, with a few non-whitespace characters. We find out the
161 -- width of this prefix, and put whitespace just as wide as that
162 -- prefix before the current line, instead of the whitespace
163 -- characters that were there.
164 -- Doing this is slightly risky, since the prefix might contain
165 -- unfinished markup (e.g., \foo{bar without the closing }). We might
166 -- need to solve this later.
167 copyind = utf.sub(indentlines[prev], 1, #newind)
168 setwidth = "\\setwidthof{" .. copyind .. "}\\to\\pretlamalignwidth"
169 hskip = "\\hskip\\pretlamalignwidth"
170 return "{" .. setwidth .. hskip .. "}" .. rest
172 -- No previous line? Just return the unmodified line then
177 -- Mark the begin of a block of lambda formatted buffers or expressions. This
178 -- means that, until you call end_of_block again, the subscript bases are
179 -- shared. For example, if you have \lam{y1} some text \lam{yn} within a
180 -- single block, the yn will properly get subscripted. Be sure to call
181 -- end_of_block again!
183 -- Blocks can be partially nested, meaning that the block
184 -- won't be closed until end_of_block was called exactly as often as
185 -- begin_of_block. However, subscripts from the inner block can still
186 -- influence subscripts in the outer block.
187 function vis.begin_of_block()
188 vis.begin_of_display()
189 in_block = in_block + 1
192 -- Ends the current block
193 function vis.end_of_block()
194 in_block = in_block - 1
197 function vis.begin_of_display()
198 if in_block == 0 then
199 -- Initially allow subscripts using _ or just appending a number (later,
200 -- we will add extra patterns here.
201 submatches = {"^(%a*)_([%a%d,]+)$", "^(%a+)([%d,]+)$"}
202 -- This stores all the bases we've encountered so far (to prevent
203 -- duplicates). For each of them there will be a pattern in submatches
211 -- Make things work for inline typeing (e.g., \type{}) as well.
212 vis.begin_of_inline = vis.begin_of_display
213 vis.end_of_inline = vis.end_of_display
215 function vis.flush_line(str,nested)
216 local result, state = { }, 0
217 local finish, change = buffers.finish_state, buffers.change_state
219 -- Set the colorscheme, which is used by finish_state and change_state
220 buffers.currentcolors = colors
224 -- See if the next token is a word
225 word, str = take_word(str)
227 if keywords[res] then
228 -- Make all keywords bold
229 word = "{\\bold " .. word .. "}"
230 if keywords[res].small then
231 word = "\\small" .. word -- Curlies were added above
234 -- Process any subscripts in the word
235 word = do_subscripts(word)
238 -- The next token is not a word, it must be a symbol
239 symbol, str = take_symbol(str)
242 -- Append the resulting token
243 result[#result+1] = word or symbol
246 state = finish(state, result)
247 buffers.flush_result(result,nested)
250 -- vim: set sw=4 sts=4 expandtab ai: