diff options
Diffstat (limited to 'lua/lexers/gnuplot.lua')
| -rw-r--r-- | lua/lexers/gnuplot.lua | 109 |
1 files changed, 45 insertions, 64 deletions
diff --git a/lua/lexers/gnuplot.lua b/lua/lexers/gnuplot.lua index 4ee72de..a3d93dd 100644 --- a/lua/lexers/gnuplot.lua +++ b/lua/lexers/gnuplot.lua @@ -1,80 +1,61 @@ --- Copyright 2006-2017 Mitchell mitchell.att.foicica.com. See LICENSE. +-- Copyright 2006-2022 Mitchell. See LICENSE. -- Gnuplot LPeg lexer. -local l = require('lexer') -local token, word_match = l.token, l.word_match -local P, R, S = lpeg.P, lpeg.R, lpeg.S +local lexer = require('lexer') +local token, word_match = lexer.token, lexer.word_match +local P, S = lpeg.P, lpeg.S -local M = {_NAME = 'gnuplot'} +local lex = lexer.new('gnuplot') -- Whitespace. -local ws = token(l.WHITESPACE, l.space^1) - --- Comments. -local comment = token(l.COMMENT, '#' * l.nonnewline^0) - --- Strings. -local sq_str = l.delimited_range("'") -local dq_str = l.delimited_range('"') -local bk_str = l.delimited_range('[]', true) -local bc_str = l.delimited_range('{}', true) -local string = token(l.STRING, sq_str + dq_str + bk_str + bc_str) +lex:add_rule('whitespace', token(lexer.WHITESPACE, lexer.space^1)) -- Keywords. -local keyword = token(l.KEYWORD, word_match{ - 'cd', 'call', 'clear', 'exit', 'fit', 'help', 'history', 'if', 'load', - 'pause', 'plot', 'using', 'with', 'index', 'every', 'smooth', 'thru', 'print', - 'pwd', 'quit', 'replot', 'reread', 'reset', 'save', 'set', 'show', 'unset', - 'shell', 'splot', 'system', 'test', 'unset', 'update' -}) +lex:add_rule('keyword', token(lexer.KEYWORD, word_match{ + 'cd', 'call', 'clear', 'exit', 'fit', 'help', 'history', 'if', 'load', 'pause', 'plot', 'using', + 'with', 'index', 'every', 'smooth', 'thru', 'print', 'pwd', 'quit', 'replot', 'reread', 'reset', + 'save', 'set', 'show', 'unset', 'shell', 'splot', 'system', 'test', 'unset', 'update' +})) -- Functions. -local func = token(l.FUNCTION, word_match{ - 'abs', 'acos', 'acosh', 'arg', 'asin', 'asinh', 'atan', 'atan2', 'atanh', - 'besj0', 'besj1', 'besy0', 'besy1', 'ceil', 'cos', 'cosh', 'erf', 'erfc', - 'exp', 'floor', 'gamma', 'ibeta', 'inverf', 'igamma', 'imag', 'invnorm', - 'int', 'lambertw', 'lgamma', 'log', 'log10', 'norm', 'rand', 'real', 'sgn', - 'sin', 'sinh', 'sqrt', 'tan', 'tanh', 'column', 'defined', 'tm_hour', - 'tm_mday', 'tm_min', 'tm_mon', 'tm_sec', 'tm_wday', 'tm_yday', 'tm_year', - 'valid' -}) +lex:add_rule('function', token(lexer.FUNCTION, word_match{ + 'abs', 'acos', 'acosh', 'arg', 'asin', 'asinh', 'atan', 'atan2', 'atanh', 'besj0', 'besj1', + 'besy0', 'besy1', 'ceil', 'cos', 'cosh', 'erf', 'erfc', 'exp', 'floor', 'gamma', 'ibeta', + 'inverf', 'igamma', 'imag', 'invnorm', 'int', 'lambertw', 'lgamma', 'log', 'log10', 'norm', + 'rand', 'real', 'sgn', 'sin', 'sinh', 'sqrt', 'tan', 'tanh', 'column', 'defined', 'tm_hour', + 'tm_mday', 'tm_min', 'tm_mon', 'tm_sec', 'tm_wday', 'tm_yday', 'tm_year', 'valid' +})) -- Variables. -local variable = token(l.VARIABLE, word_match{ - 'angles', 'arrow', 'autoscale', 'bars', 'bmargin', 'border', 'boxwidth', - 'clabel', 'clip', 'cntrparam', 'colorbox', 'contour', 'datafile ', - 'decimalsign', 'dgrid3d', 'dummy', 'encoding', 'fit', 'fontpath', 'format', - 'functions', 'function', 'grid', 'hidden3d', 'historysize', 'isosamples', - 'key', 'label', 'lmargin', 'loadpath', 'locale', 'logscale', 'mapping', - 'margin', 'mouse', 'multiplot', 'mx2tics', 'mxtics', 'my2tics', 'mytics', - 'mztics', 'offsets', 'origin', 'output', 'parametric', 'plot', 'pm3d', - 'palette', 'pointsize', 'polar', 'print', 'rmargin', 'rrange', 'samples', - 'size', 'style', 'surface', 'terminal', 'tics', 'ticslevel', 'ticscale', - 'timestamp', 'timefmt', 'title', 'tmargin', 'trange', 'urange', 'variables', - 'version', 'view', 'vrange', 'x2data', 'x2dtics', 'x2label', 'x2mtics', - 'x2range', 'x2tics', 'x2zeroaxis', 'xdata', 'xdtics', 'xlabel', 'xmtics', - 'xrange', 'xtics', 'xzeroaxis', 'y2data', 'y2dtics', 'y2label', 'y2mtics', - 'y2range', 'y2tics', 'y2zeroaxis', 'ydata', 'ydtics', 'ylabel', 'ymtics', - 'yrange', 'ytics', 'yzeroaxis', 'zdata', 'zdtics', 'cbdata', 'cbdtics', - 'zero', 'zeroaxis', 'zlabel', 'zmtics', 'zrange', 'ztics', 'cblabel', - 'cbmtics', 'cbrange', 'cbtics' -}) +lex:add_rule('variable', token(lexer.VARIABLE, word_match{ + 'angles', 'arrow', 'autoscale', 'bars', 'bmargin', 'border', 'boxwidth', 'clabel', 'clip', + 'cntrparam', 'colorbox', 'contour', 'datafile', 'decimalsign', 'dgrid3d', 'dummy', 'encoding', + 'fit', 'fontpath', 'format', 'functions', 'function', 'grid', 'hidden3d', 'historysize', + 'isosamples', 'key', 'label', 'lmargin', 'loadpath', 'locale', 'logscale', 'mapping', 'margin', + 'mouse', 'multiplot', 'mx2tics', 'mxtics', 'my2tics', 'mytics', 'mztics', 'offsets', 'origin', + 'output', 'parametric', 'plot', 'pm3d', 'palette', 'pointsize', 'polar', 'print', 'rmargin', + 'rrange', 'samples', 'size', 'style', 'surface', 'terminal', 'tics', 'ticslevel', 'ticscale', + 'timestamp', 'timefmt', 'title', 'tmargin', 'trange', 'urange', 'variables', 'version', 'view', + 'vrange', 'x2data', 'x2dtics', 'x2label', 'x2mtics', 'x2range', 'x2tics', 'x2zeroaxis', 'xdata', + 'xdtics', 'xlabel', 'xmtics', 'xrange', 'xtics', 'xzeroaxis', 'y2data', 'y2dtics', 'y2label', + 'y2mtics', 'y2range', 'y2tics', 'y2zeroaxis', 'ydata', 'ydtics', 'ylabel', 'ymtics', 'yrange', + 'ytics', 'yzeroaxis', 'zdata', 'zdtics', 'cbdata', 'cbdtics', 'zero', 'zeroaxis', 'zlabel', + 'zmtics', 'zrange', 'ztics', 'cblabel', 'cbmtics', 'cbrange', 'cbtics' +})) -- Identifiers. -local identifier = token(l.IDENTIFIER, l.word) +lex:add_rule('identifier', token(lexer.IDENTIFIER, lexer.word)) --- Operators. -local operator = token(l.OPERATOR, S('-+~!$*%=<>&|^?:()')) +-- Strings. +local sq_str = lexer.range("'") +local dq_str = lexer.range('"') +local br_str = lexer.range('[', ']', true) + lexer.range('{', '}', true) +lex:add_rule('string', token(lexer.STRING, sq_str + dq_str + br_str)) + +-- Comments. +lex:add_rule('comment', token(lexer.COMMENT, lexer.to_eol('#'))) -M._rules = { - {'whitespace', ws}, - {'keyword', keyword}, - {'function', func}, - {'variable', variable}, - {'identifier', identifier}, - {'string', string}, - {'comment', comment}, - {'operator', operator}, -} +-- Operators. +lex:add_rule('operator', token(lexer.OPERATOR, S('-+~!$*%=<>&|^?:()'))) -return M +return lex |
