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