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 [' '] = {repr = '\\obs '},
21 ['_'] = {repr = '\\_'},
22 ['->'] = {repr = '\\rightarrow'},
23 -- The default * sits very high above the baseline, \ast (u+2217) looks
25 ['*'] = {repr = '\\ast'},
26 ['~'] = {repr = '\\HDLine[width=.20 * \\the\\textwidth]'},
27 ['|'] = {repr = '\\char' .. utf.byte('|')},
30 -- Keywords that should be bold
44 -- Store the last line for each indent level
45 local indentlines = {}
47 -- See if str starts with a symbol, and return the remaining string and that
48 -- symbol. If no symbol from the table is matched, just returns the first
49 -- character. We can do a lookup directly, since symbols can be different in
50 -- length, so we just loop over all symbols, trying them in turn.
51 local function take_symbol(str)
52 for symbol,props in pairs(symbols) do
53 -- Try to remove symbol from the start of str
54 symbol, newstr = utf.match(str, "^(" .. symbol .. ")(.*)")
56 -- Return this tokens repr, or just the token if it has no
58 res = props.repr or symbol
59 -- Enclose the token in {\style .. }
61 res = "{\\" .. props.style .. " " .. res .. "}"
66 -- No symbol found, just return the first character
67 return utf.match(str, "^(.)(.*)")
70 -- Take a single word from str, if posible. Returns the rest of the string and
72 local function take_word(str)
73 -- A word must always start with a-z (in particular, λ is not a valid
75 res, newstr = utf.match(str, "^([a-zA-Z][%a%d%+%-%,_]+)(.*)")
76 return res, newstr or str
79 -- Tries to match each of the patterns and returns the captures of the first
80 -- matching pattern (up to 5 captures are supported). Returns nil when nothing
82 local function match_mul(str, patterns)
83 for i, pat in ipairs(patterns) do
84 a, b, c, d, e = utf.match(str, pat)
92 -- Find any subscripts in the given word and typeset them
93 local function do_subscripts(word)
94 base, sub = match_mul(res, submatches)
96 word = base .. "\\low{" .. sub .. "}"
97 -- After a word has been used as a base, allow subscripts
98 -- without _, even for non-numbers.
99 if not bases[base] then
100 -- Register that we've added this base
102 -- Add a patterns for this base. First, the base with a single
103 -- letter or number subscript.
104 submatches[#submatches+1] = "^(" .. base .. ")([%a%d])$"
105 -- Seconde, the base with a longer prefix that includes at least
106 -- one of +-, (to catch things like ri+1, but not return).
107 submatches[#submatches+1] = "^(" .. base .. ")([%a%d]*[%-%+%,]+[%a%d%-%+%,]*)$"
113 -- Do proper aligning for subsequent lines. For example, in
116 -- We replace the spaces in the second line with a skip with the same with as
117 -- "foo ", to align the | with the =.
118 -- For this, we keep a table "indentlines", which contains all previous lines
119 -- with smaller indent levels that are still "in scope" (e.g., have not yet
120 -- been followed by a line with a smaller indent level). For example:
126 -- After the last line, the table will contain:
127 -- { 0 = "line1", 2 = " line4", 4 = " line5"}
128 -- In other words, line3 is no longer in scope since it is "hidden" by
129 -- line4, and line is no longer in scope since it is replaced by line4.
130 local function do_indent(line)
131 newind, rest = utf.match(line, '^(%s*)(.*)')
133 -- Loop all the previous lines
134 for indent, unused in pairs(indentlines) do
135 if indent > #newind then
136 -- Remove any lines with a larger indent
137 indentlines[indent] = nil
138 elseif indent < #newind and indent > prev then
139 -- Find the last line (e.g, with the highest indent) with an
140 -- indent smaller than the new indent. This is the line from which
141 -- we need to copy the indent.
146 -- Always store this line, possibly overwriting a previous line with the
148 indentlines[#newind] = line
151 -- If there is a previous line with a smaller indent, make sure we
152 -- align with it. We do this by taking a prefix from that previous
153 -- line just as long as our indent. This gives us a bunch of
154 -- whitespace, with a few non-whitespace characters. We find out the
155 -- width of this prefix, and put whitespace just as wide as that
156 -- prefix before the current line, instead of the whitespace
157 -- characters that were there.
158 -- Doing this is slightly risky, since the prefix might contain
159 -- unfinished markup (e.g., \foo{bar without the closing }). We might
160 -- need to solve this later.
161 copyind = utf.sub(indentlines[prev], 1, #newind)
162 setwidth = "\\setwidthof{" .. copyind .. "}\\to\\pretlamalignwidth"
163 hskip = "\\hskip\\pretlamalignwidth"
164 return "{" .. setwidth .. hskip .. "}" .. rest
166 -- No previous line? Just return the unmodified line then
171 -- Mark the begin of a block of lambda formatted buffers or expressions. This
172 -- means that, until you call end_of_block again, the subscript bases are
173 -- shared. For example, if you have \lam{y1} some text \lam{yn} within a
174 -- single block, the yn will properly get subscripted. Be sure to call
175 -- end_of_block again!
177 -- Blocks can be partially nested, meaning that the block
178 -- won't be closed until end_of_block was called exactly as often as
179 -- begin_of_block. However, subscripts from the inner block can still
180 -- influence subscripts in the outer block.
181 function vis.begin_of_block()
182 vis.begin_of_display()
183 in_block = in_block + 1
186 -- Ends the current block
187 function vis.end_of_block()
188 in_block = in_block - 1
191 function vis.begin_of_display()
192 if in_block == 0 then
193 -- Initially allow subscripts using _ or just appending a number (later,
194 -- we will add extra patterns here.
195 submatches = {"^(%a*)_([%a%d,]+)$", "^(%a+)([%d,]+)$"}
196 -- This stores all the bases we've encountered so far (to prevent
197 -- duplicates). For each of them there will be a pattern in submatches
205 -- Make things work for inline typeing (e.g., \type{}) as well.
206 vis.begin_of_inline = vis.begin_of_display
207 vis.end_of_inline = vis.end_of_display
209 function vis.flush_line(str,nested)
210 local result, state = { }, 0
211 local finish, change = buffers.finish_state, buffers.change_state
213 -- Set the colorscheme, which is used by finish_state and change_state
214 buffers.currentcolors = colors
218 -- See if the next token is a word
219 word, str = take_word(str)
221 if keywords[res] then
222 -- Make all keywords bold
223 word = "{\\bold " .. word .. "}"
225 -- Process any subscripts in the word
226 word = do_subscripts(word)
229 -- The next token is not a word, it must be a symbol
230 symbol, str = take_symbol(str)
233 -- Append the resulting token
234 result[#result+1] = word or symbol
237 state = finish(state, result)
238 buffers.flush_result(result,nested)
241 -- vim: set sw=4 sts=4 expandtab ai: