lexers/perl.lua

   1 -- Copyright 2006-2015 Mitchell mitchell.att.foicica.com. See LICENSE.
   2 -- Perl LPeg lexer.
   3
   4 local l = require('lexer')
   5 local token, word_match = l.token, l.word_match
   6 local P, R, S, V = lpeg.P, lpeg.R, lpeg.S, lpeg.V
   7
   8 local M = {_NAME = 'perl'}
   9
  10 -- Whitespace.
  11 local ws = token(l.WHITESPACE, l.space^1)
  12
  13 -- Comments.
  14 local line_comment = '#' * l.nonnewline_esc^0
  15 local block_comment = l.starts_line('=') * l.alpha *
  16                       (l.any - l.newline * '=cut')^0 * (l.newline * '=cut')^-1
  17 local comment = token(l.COMMENT, block_comment + line_comment)
  18
  19 local delimiter_matches = {['('] = ')', ['['] = ']', ['{'] = '}', ['<'] = '>'}
  20 local literal_delimitted = P(function(input, index) -- for single delimiter sets
  21   local delimiter = input:sub(index, index)
  22   if not delimiter:find('%w') then -- only non alpha-numerics
  23     local match_pos, patt
  24     if delimiter_matches[delimiter] then
  25       -- Handle nested delimiter/matches in strings.
  26       local s, e = delimiter, delimiter_matches[delimiter]
  27       patt = l.delimited_range(s..e, false, false, true)
  28     else
  29       patt = l.delimited_range(delimiter)
  30     end
  31     match_pos = lpeg.match(patt, input, index)
  32     return match_pos or #input + 1
  33   end
  34 end)
  35 local literal_delimitted2 = P(function(input, index) -- for 2 delimiter sets
  36   local delimiter = input:sub(index, index)
  37   -- Only consider non-alpha-numerics and non-spaces as delimiters. The
  38   -- non-spaces are used to ignore operators like "-s".
  39   if not delimiter:find('[%w ]') then
  40     local match_pos, patt
  41     if delimiter_matches[delimiter] then
  42       -- Handle nested delimiter/matches in strings.
  43       local s, e = delimiter, delimiter_matches[delimiter]
  44       patt = l.delimited_range(s..e, false, false, true)
  45     else
  46       patt = l.delimited_range(delimiter)
  47     end
  48     first_match_pos = lpeg.match(patt, input, index)
  49     final_match_pos = lpeg.match(patt, input, first_match_pos - 1)
  50     if not final_match_pos then -- using (), [], {}, or <> notation
  51       final_match_pos = lpeg.match(l.space^0 * patt, input, first_match_pos)
  52     end
  53     return final_match_pos or #input + 1
  54   end
  55 end)
  56
  57 -- Strings.
  58 local sq_str = l.delimited_range("'")
  59 local dq_str = l.delimited_range('"')
  60 local cmd_str = l.delimited_range('`')
  61 local heredoc = '<<' * P(function(input, index)
  62   local s, e, delimiter = input:find('([%a_][%w_]*)[\n\r\f;]+', index)
  63   if s == index and delimiter then
  64     local end_heredoc = '[\n\r\f]+'
  65     local _, e = input:find(end_heredoc..delimiter, e)
  66     return e and e + 1 or #input + 1
  67   end
  68 end)
  69 local lit_str = 'q' * P('q')^-1 * literal_delimitted
  70 local lit_array = 'qw' * literal_delimitted
  71 local lit_cmd = 'qx' * literal_delimitted
  72 local lit_tr = (P('tr') + 'y') * literal_delimitted2 * S('cds')^0
  73 local regex_str = #P('/') * l.last_char_includes('-<>+*!~\\=%&|^?:;([{') *
  74                   l.delimited_range('/', true) * S('imosx')^0
  75 local lit_regex = 'qr' * literal_delimitted * S('imosx')^0
  76 local lit_match = 'm' * literal_delimitted * S('cgimosx')^0
  77 local lit_sub = 's' * literal_delimitted2 * S('ecgimosx')^0
  78 local string = token(l.STRING, sq_str + dq_str + cmd_str + heredoc + lit_str +
  79                                lit_array + lit_cmd + lit_tr) +
  80                token(l.REGEX, regex_str + lit_regex + lit_match + lit_sub)
  81
  82 -- Numbers.
  83 local number = token(l.NUMBER, l.float + l.integer)
  84
  85 -- Keywords.
  86 local keyword = token(l.KEYWORD, word_match{
  87   'STDIN', 'STDOUT', 'STDERR', 'BEGIN', 'END', 'CHECK', 'INIT',
  88   'require', 'use',
  89   'break', 'continue', 'do', 'each', 'else', 'elsif', 'foreach', 'for', 'if',
  90   'last', 'local', 'my', 'next', 'our', 'package', 'return', 'sub', 'unless',
  91   'until', 'while', '__FILE__', '__LINE__', '__PACKAGE__',
  92   'and', 'or', 'not', 'eq', 'ne', 'lt', 'gt', 'le', 'ge'
  93 })
  94
  95 -- Functions.
  96 local func = token(l.FUNCTION, word_match({
  97   'abs', 'accept', 'alarm', 'atan2', 'bind', 'binmode', 'bless', 'caller',
  98   'chdir', 'chmod', 'chomp', 'chop', 'chown', 'chr', 'chroot', 'closedir',
  99   'close', 'connect', 'cos', 'crypt', 'dbmclose', 'dbmopen', 'defined',
 100   'delete', 'die', 'dump', 'each', 'endgrent', 'endhostent', 'endnetent',
 101   'endprotoent', 'endpwent', 'endservent', 'eof', 'eval', 'exec', 'exists',
 102   'exit', 'exp', 'fcntl', 'fileno', 'flock', 'fork', 'format', 'formline',
 103   'getc', 'getgrent', 'getgrgid', 'getgrnam', 'gethostbyaddr', 'gethostbyname',
 104   'gethostent', 'getlogin', 'getnetbyaddr', 'getnetbyname', 'getnetent',
 105   'getpeername', 'getpgrp', 'getppid', 'getpriority', 'getprotobyname',
 106   'getprotobynumber', 'getprotoent', 'getpwent', 'getpwnam', 'getpwuid',
 107   'getservbyname', 'getservbyport', 'getservent', 'getsockname', 'getsockopt',
 108   'glob', 'gmtime', 'goto', 'grep', 'hex', 'import', 'index', 'int', 'ioctl',
 109   'join', 'keys', 'kill', 'lcfirst', 'lc', 'length', 'link', 'listen',
 110   'localtime', 'log', 'lstat', 'map', 'mkdir', 'msgctl', 'msgget', 'msgrcv',
 111   'msgsnd', 'new', 'oct', 'opendir', 'open', 'ord', 'pack', 'pipe', 'pop',
 112   'pos', 'printf', 'print', 'prototype', 'push', 'quotemeta', 'rand', 'readdir',
 113   'read', 'readlink', 'recv', 'redo', 'ref', 'rename', 'reset', 'reverse',
 114   'rewinddir', 'rindex', 'rmdir', 'scalar', 'seekdir', 'seek', 'select',
 115   'semctl', 'semget', 'semop', 'send', 'setgrent', 'sethostent', 'setnetent',
 116   'setpgrp', 'setpriority', 'setprotoent', 'setpwent', 'setservent',
 117   'setsockopt', 'shift', 'shmctl', 'shmget', 'shmread', 'shmwrite', 'shutdown',
 118   'sin', 'sleep', 'socket', 'socketpair', 'sort', 'splice', 'split', 'sprintf',
 119   'sqrt', 'srand', 'stat', 'study', 'substr', 'symlink', 'syscall', 'sysread',
 120   'sysseek', 'system', 'syswrite', 'telldir', 'tell', 'tied', 'tie', 'time',
 121   'times', 'truncate', 'ucfirst', 'uc', 'umask', 'undef', 'unlink', 'unpack',
 122   'unshift', 'untie', 'utime', 'values', 'vec', 'wait', 'waitpid', 'wantarray',
 123   'warn', 'write'
 124 }, '2'))
 125
 126 -- Identifiers.
 127 local identifier = token(l.IDENTIFIER, l.word)
 128
 129 -- Variables.
 130 local special_var = '$' * ('^' * S('ADEFHILMOPSTWX')^-1 +
 131                            S('\\"[]\'&`+*.,;=%~?@<>(|/!-') +
 132                            ':' * (l.any - ':') + P('$') * -l.word + l.digit^1)
 133 local plain_var = ('$#' + S('$@%')) * P('$')^0 * l.word + '$#'
 134 local variable = token(l.VARIABLE, special_var + plain_var)
 135
 136 -- Operators.
 137 local operator = token(l.OPERATOR, S('-<>+*!~\\=/%&|^.?:;()[]{}'))
 138
 139 -- Markers.
 140 local marker = token(l.COMMENT, word_match{'__DATA__', '__END__'} * l.any^0)
 141
 142 M._rules = {
 143   {'whitespace', ws},
 144   {'keyword', keyword},
 145   {'marker', marker},
 146   {'function', func},
 147   {'string', string},
 148   {'identifier', identifier},
 149   {'comment', comment},
 150   {'number', number},
 151   {'variable', variable},
 152   {'operator', operator},
 153 }
 154
 155 M._foldsymbols = {
 156   _patterns = {'[%[%]{}]', '#'},
 157   [l.OPERATOR] = {['['] = 1, [']'] = -1, ['{'] = 1, ['}'] = -1},
 158   [l.COMMENT] = {['#'] = l.fold_line_comments('#')}
 159 }
 160
 161 return M