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