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