fix: syntax and ftplugin fixes

README.md

@@ -32,8 +32,8 @@ characters they both highlight.
   values, plot styles, plot-element modifiers, terminal names, built-in
   functions and variables, datablocks, macros, numbers with unit suffixes,
   and both string flavours.
-- Sensible buffer-local editor settings (comments, `iskeyword`,
-  `matchpairs`), all undone by `b:undo_ftplugin`.
+- Sensible buffer-local editor settings (comments, `iskeyword`), all undone
+  by `b:undo_ftplugin`.
 - No commands, no mappings, no global options. Nothing to configure.
 - Pure Vim script.
 
@@ -110,8 +110,6 @@ The ftplugin sets a few buffer-local options:
   code does not auto-wrap.
 - `iskeyword+=_`, because gnuplot identifiers and its built-in variables
   (`GPVAL_TERM`, `STATS_mean_y`) contain underscores.
-- `matchpairs+=<:>`, so `%` jumps between the brackets in
-  `set xrange [0:1]` style pairs.
 
 All of it is registered in `b:undo_ftplugin`, so `:setfiletype` away from
 gnuplot restores your defaults.

doc/gnuplot.txt

@@ -59,9 +59,6 @@ The ftplugin sets a few buffer-local options:
     'iskeyword' gains `_`, because gnuplot identifiers and its built-in
     variables (`GPVAL_TERM`, `STATS_mean_y`) contain underscores.
 
-    'matchpairs' gains `<:>`, so |%| jumps between the brackets in
-    `set xrange [0:1]` style pairs.
-
 Everything is registered in |b:undo_ftplugin|, so leaving the filetype
 restores your defaults.
 

ftplugin/gnuplot.vim

@@ -21,11 +21,8 @@ setlocal formatoptions+=croql
 " contain underscores.
 setlocal iskeyword+=_
 
-" `set xrange [0:1]` and friends: treat brackets as matched pairs.
-setlocal matchpairs+=<:>
-
 let b:undo_ftplugin = 'setlocal commentstring< comments< formatoptions<'
-      \ . ' iskeyword< matchpairs<'
+      \ . ' iskeyword<'
 
 let &cpoptions = s:cpo_save
 unlet s:cpo_save

syntax/gnuplot.vim

@@ -26,20 +26,27 @@ set cpoptions&vim
 syn iskeyword @,48-57,_
 
 " Comments ------------------------------------------------------------------
-" A trailing backslash continues a comment onto the next line.
-syn match   gnuplotComment "#.*\%(\\\n.*\)*$" contains=gnuplotTodo,@Spell
+" A trailing backslash continues a comment onto the next line, so the comment
+" ends only at an end of line that is not preceded by one.
+syn region  gnuplotComment start="#" end="\\\@<!$" contains=gnuplotTodo,@Spell
 syn keyword gnuplotTodo contained TODO FIXME XXX BUG NOTE HACK
 
 " Strings -------------------------------------------------------------------
-" Double quotes interpret backslash escapes; single quotes do not.
+" Double quotes interpret backslash escapes; single quotes do not. gnuplot
+" closes an unterminated string at the end of the line unless a trailing
+" backslash continues it.
 syn match   gnuplotEscape contained "\\\%([\\abfnrtv"']\|[0-7]\{1,3}\|x\x\{1,2}\|u\x\{4}\|U\x\{8}\)"
 syn match   gnuplotFormat  contained "%[-+ #0]*\d*\%(\.\d\+\)\?[hlLqjzt]*[diouxXeEfgGcsSpP%]"
-syn region  gnuplotString  start=+"+ skip=+\\.+ end=+"+ contains=gnuplotEscape,gnuplotFormat,@Spell
-syn region  gnuplotString  start=+'+ end=+'+ contains=gnuplotFormat,@Spell
+syn region  gnuplotString  start=+"+ skip=+\\.+ end=+"\|\\\@<!$+ contains=gnuplotEscape,gnuplotFormat,@Spell
+syn region  gnuplotString  start=+'+ end=+'\|\\\@<!$+ contains=gnuplotFormat,@Spell
 
 " Datablocks ----------------------------------------------------------------
 " `$data << EOD ... EOD` — the terminator is user-chosen, so match the whole
-" block generically rather than pinning it to EOD.
+" block generically rather than pinning it to EOD. The header may follow
+" another command on the same line (`print 1; $d << EOD`), and `function $f <<
+" EOF` blocks take the same form. gnuplot ends the block at the first line that
+" begins with the terminator in column one, so `  EOD` is still data and
+" `EODxyz` closes it.
 " Column references in a using spec (`using ($1+$2)`). Digits, so this can
 " never collide with the `$name` forms below.
 syn match   gnuplotColumnRef "\$\d\+"
@@ -49,7 +56,7 @@ syn match   gnuplotColumnRef "\$\d\+"
 " on `$name` would out-rank the region and stop it opening at all.
 syn match   gnuplotDatablockRef "\$\h\w*"
 syn region  gnuplotDatablock matchgroup=gnuplotDatablockName contains=NONE
-      \ start="^\s*\$\h\w*\s*<<\s*\z(\h\w*\)" end="^\s*\z1\s*$"
+      \ start="\$\h\w*\s*<<\s*\z(\h\w*\)" end="^\z1"
 syn match   gnuplotOperator "<<" containedin=gnuplotDatablockName
 
 " Macros --------------------------------------------------------------------
@@ -58,9 +65,11 @@ syn match   gnuplotMacro "@\h\w*"
 " Numbers -------------------------------------------------------------------
 " gnuplot folds a unit suffix into the number itself (10cm, 3.0in, 0.5pt). The
 " full set is the grammar's typographic-unit alternation; `inch` has to precede
-" `in` or the longer spelling never matches.
-syn match   gnuplotNumber "\<\d\+\%(\.\d*\)\?\%([eE][-+]\?\d\+\)\?\%(cm\|inch\|in\|mm\|pt\|pc\|bp\|dd\|cc\|pi\)\?\>"
-syn match   gnuplotNumber "\<\.\d\+\%([eE][-+]\?\d\+\)\?\%(cm\|inch\|in\|mm\|pt\|pc\|bp\|dd\|cc\|pi\)\?\>"
+" `in` or the longer spelling never matches. A trailing dot (`1.`) belongs to
+" the number, and gnuplot reads a dot before a digit as a number start (`.5`)
+" even right after an identifier.
+syn match   gnuplotNumber "\<\d\+\%(\.\d*\)\?\%([eE][-+]\?\d\+\)\?\%(cm\|inch\|in\|mm\|pt\|pc\|bp\|dd\|cc\|pi\)\?\w\@!"
+syn match   gnuplotNumber "\.\d\+\%([eE][-+]\?\d\+\)\?\%(cm\|inch\|in\|mm\|pt\|pc\|bp\|dd\|cc\|pi\)\?\>"
 syn match   gnuplotNumber "\<0[xX]\x\+\>"
 
 " Built-in variables --------------------------------------------------------
@@ -75,18 +84,19 @@ syn match   gnuplotBuiltinVar "\<\w\+_\%(records\|headers\|outofrange\|invalid\|
 syn keyword gnuplotBuiltinVar GNUTERM VoxelDistance GridDistance
 
 " Functions -----------------------------------------------------------------
-" Any identifier immediately followed by '(' reads as a call; the built-in list
+" Any identifier followed by '(' reads as a call; the built-in list
 " is defined afterwards so it wins on the same match position.
-syn match   gnuplotFunction "\<\h\w*\ze("
-syn match   gnuplotBuiltinFunction "\<\%(abs\|acosh\|acos\|airy\|arg\|asinh\|asin\|atan2\|atanh\|atan\|cbrt\|ceil\|conj\|cosh\|cos\|exp\|floor\|imag\|int\|log10\|log\|norm\|rand\|real\|round\|sgn\|sinh\|sin\|sqrt\|tanh\|tan\)\ze("
-syn match   gnuplotBuiltinFunction "\<\%(bes[ijy][01n]\|Ai\|Bi\|Bessel[HIJKY]1\?2\?\|Elliptic[EKP]i\?\|erfc\|erfi\|erf\|expint\|gamma\|ibeta\|igamma\|inverf\|invibeta\|invigamma\|invnorm\|LambertW\|lgamma\|lnGamma\|SynchrotronF\|uigamma\|voigt\|zeta\)\ze("
-syn match   gnuplotBuiltinFunction "\<\%(cdawson\|cerf\|faddeeva\|Fresnel[CS]\|VP\%(_fwhm\)\?\)\ze("
-syn match   gnuplotBuiltinFunction "\<\%(gprintf\|sprintf\|strlen\|strstrt\|substr\|split\|join\|trim\|word\|words\|system\)\ze("
-syn match   gnuplotBuiltinFunction "\<\%(column\|columnhead\%(er\)\?\|stringcolumn\|strcol\|exists\|valid\|value\|hsv2rgb\|palette\|rgbcolor\|voxel\|index\)\ze("
-syn match   gnuplotBuiltinFunction "\<\%(time\%(column\)\?\|strftime\|strptime\|tm_\w\+\|weekdate_\%(iso\|cdc\)\)\ze("
+syn match   gnuplotFunction "\<\h\w*\ze\s*("
+syn match   gnuplotBuiltinFunction "\<\%(abs\|acosh\|acos\|airy\|arg\|asinh\|asin\|atan2\|atanh\|atan\|cbrt\|ceil\|conj\|cosh\|cos\|exp\|floor\|imag\|int\|log10\|log\|norm\|rand\|real\|round\|sgn\|sinh\|sin\|sqrt\|tanh\|tan\)\ze\s*("
+syn match   gnuplotBuiltinFunction "\<\%(bes[ijy][01n]\|Ai\|Bi\|Bessel[HIJKY]1\?2\?\|Elliptic[EKP]i\?\|erfc\|erfi\|erf\|expint\|gamma\|ibeta\|igamma\|inverf\|invibeta\|invigamma\|invnorm\|LambertW\|lgamma\|lnGamma\|SynchrotronF\|uigamma\|voigt\|zeta\)\ze\s*("
+syn match   gnuplotBuiltinFunction "\<\%(cdawson\|cerf\|faddeeva\|Fresnel[CS]\|VP\%(_fwhm\)\?\)\ze\s*("
+syn match   gnuplotBuiltinFunction "\<\%(gprintf\|sprintf\|strlen\|strstrt\|substr\|split\|join\|trim\|word\|words\|system\)\ze\s*("
+syn match   gnuplotBuiltinFunction "\<\%(column\|columnhead\%(er\)\?\|stringcolumn\|strcol\|exists\|valid\|value\|hsv2rgb\|palette\|rgbcolor\|voxel\|index\)\ze\s*("
+syn match   gnuplotBuiltinFunction "\<\%(time\%(column\)\?\|strftime\|strptime\|tm_\w\+\|weekdate_\%(iso\|cdc\)\)\ze\s*("
 
 " Operators -----------------------------------------------------------------
-syn match   gnuplotOperator "[-+*/%^!~?:.,]"
+" A dot before a digit starts a number, not a concatenation.
+syn match   gnuplotOperator "[-+*/%^!~?:,]\|\.\d\@!"
 syn match   gnuplotOperator "\*\*"
 syn match   gnuplotOperator "[=!<>]="
 syn match   gnuplotOperator "[<>]"
@@ -364,6 +374,11 @@ syn keyword gnuplotConditional if else
 syn keyword gnuplotRepeat      do while for in
 syn keyword gnuplotCommand     array sum
 
+" Synchronization -----------------------------------------------------------
+" Datablocks have no length limit, so no fixed lookback can tell whether a line
+" sits inside one. Parse from the top of the file instead.
+syn sync fromstart
+
 " Highlight links -----------------------------------------------------------
 " The group names come from the grammar's capture taxonomy, where an alias
 " names the clause family a word belongs to rather than the word itself: `cmd`

tests/run.vim

@@ -13,6 +13,9 @@ filetype plugin on
 syntax enable
 
 let v:errors = []
+" A script error (bad expression, unknown function) skips its assertion
+" without touching v:errors, and nvim still exits 0, so it is caught below.
+let v:errmsg = ''
 
 " Raw syntax group at (line, col). Deliberately NOT synIDtrans: Vim collapses
 " Conditional, Repeat and Keyword all into Statement, which would make most of
@@ -35,6 +38,22 @@ function! s:Word(text, word, expected) abort
         \ printf('%s  ->  %s', a:text, a:word))
 endfunction
 
+" Assert the group under the first occurrence of {word} on line {lnum} of a
+" multi-line buffer. Queried bottom-up, so the answer depends on the file
+" syncing from the top rather than on what was parsed before.
+function! s:Line(lines, lnum, word, expected) abort
+  enew!
+  setlocal filetype=gnuplot
+  call setline(1, a:lines)
+  let l:col = stridx(a:lines[a:lnum - 1], a:word) + 1
+  if l:col <= 0
+    call add(v:errors, printf('fixture error: %s not on line %d', a:word, a:lnum))
+    return
+  endif
+  call assert_equal(a:expected, s:Group(a:lnum, l:col),
+        \ printf('%s  ->  line %d %s', join(a:lines, ' | '), a:lnum, a:word))
+endfunction
+
 " --- commands ---------------------------------------------------------------
 call s:Word('plot sin(x)', 'plot', 'gnuplotCommand')
 call s:Word('p sin(x)', 'p', 'gnuplotCommand')
@@ -92,6 +111,17 @@ call s:Word('set size 1inch, 2mm', '1inch', 'gnuplotNumber')
 
 " datablock contents are inert
 call s:Word('$d << EOD', '$d', 'gnuplotDatablockName')
+" A deep line reads as data even when nothing above it was parsed first.
+call s:Line(['$d << EOD'] + repeat(['plot 1 # x'], 200) + ['EOD'], 150,
+      \ 'plot', 'gnuplotDatablock')
+" gnuplot ends the block on the first line starting with the terminator in
+" column one: `  EOD` is data, `EODxyz` closes the block.
+call s:Line(['$d << EOD', '  EOD', 'plot 1'], 2, 'EOD', 'gnuplotDatablock')
+call s:Line(['$d << EOD', '  EOD', 'plot 1'], 3, 'plot', 'gnuplotDatablock')
+call s:Line(['$d << EOD', '1 2', 'EODxyz', 'plot 1'], 4, 'plot', 'gnuplotCommand')
+" The header may follow another command on the same line.
+call s:Line(['print 1; $d << EOD', '1 2', 'EOD'], 1, '$d', 'gnuplotDatablockName')
+call s:Line(['print 1; $d << EOD', '1 2', 'EOD'], 2, '1', 'gnuplotDatablock')
 call s:Word('set key noautotitle', 'noautotitle', 'gnuplotOption')
 
 " --- tiers ------------------------------------------------------------------
@@ -105,7 +135,24 @@ call s:Word('set border front', 'front', 'gnuplotFlag')
 call s:Word('plot "data.dat"', '"data.dat"', 'gnuplotString')
 call s:Word("plot 'data.dat'", "'data.dat'", 'gnuplotString')
 call s:Word('# a comment', '# a comment', 'gnuplotComment')
+" A trailing backslash continues the comment; without one the next line is code.
+call s:Line(['# comment\', 'print 42'], 2, 'print', 'gnuplotComment')
+call s:Line(['print 1 # c\', 'print 42'], 2, '42', 'gnuplotComment')
+call s:Line(['# comment', 'print 42'], 2, 'print', 'gnuplotCommand')
+" gnuplot closes an unterminated string at the end of the line, unless a
+" trailing backslash continues it.
+call s:Line(['print "unfinished', 'print 42'], 2, 'print', 'gnuplotCommand')
+call s:Line(["print 'unfinished", 'print 42'], 2, 'print', 'gnuplotCommand')
+call s:Line(['print "abc\', 'def"', 'print 4'], 2, 'def', 'gnuplotString')
+call s:Line(['print "abc\', 'def"', 'print 4'], 3, 'print', 'gnuplotCommand')
+call s:Line(['print ''abc\', 'def''', 'print 4'], 2, 'def', 'gnuplotString')
+call s:Word('print "a\"b" . x', 'x', '')
 call s:Word('a = 42', '42', 'gnuplotNumber')
+" Leading and trailing dots are part of the number, not the `.` operator.
+call s:Word('a = .5', '.', 'gnuplotNumber')
+call s:Word('a = 1.', '.', 'gnuplotNumber')
+call s:Word('a = 1.e3', 'e', 'gnuplotNumber')
+call s:Word('a = "x" . b', '.', 'gnuplotOperator')
 call s:Word('set size 10cm, 5cm', '10cm', 'gnuplotNumber')
 call s:Word('a = pi', 'pi', 'gnuplotConstant')
 call s:Word('load @macro', '@macro', 'gnuplotMacro')
@@ -115,12 +162,22 @@ call s:Word('print STATS_mean_y', 'STATS_mean_y', 'gnuplotBuiltinVar')
 " --- functions --------------------------------------------------------------
 call s:Word('plot sin(x)', 'sin', 'gnuplotBuiltinFunction')
 call s:Word('a = myfunc(2)', 'myfunc', 'gnuplotFunction')
+" gnuplot allows whitespace before the argument list.
+call s:Word('print sin (0)', 'sin', 'gnuplotBuiltinFunction')
+call s:Word('a = myfunc (2)', 'myfunc', 'gnuplotFunction')
 
 " --- control flow -----------------------------------------------------------
 call s:Word('if (a > 1) { print "y" }', 'if', 'gnuplotConditional')
 call s:Word('do for [i=1:10] { print i }', 'do', 'gnuplotRepeat')
 call s:Word('do for [i=1:10] { print i }', 'for', 'gnuplotRepeat')
 
+" --- ftplugin ---------------------------------------------------------------
+" `<` and `>` are comparisons in gnuplot, never delimiters, so `%` must not
+" pair them.
+enew!
+setlocal filetype=gnuplot
+call assert_equal(-1, stridx(&l:matchpairs, '<:>'), 'matchpairs has no <:>')
+
 " --- filetype detection -----------------------------------------------------
 " `.gp` is not claimed: it belongs to PARI/GP in the stock runtimes.
 for s:name in ['t.plt', 't.plot', 't.gnu', 't.gpi', 't.gnuplot',
@@ -150,6 +207,10 @@ function! s:Report(lines) abort
   endif
 endfunction
 
+if !empty(v:errmsg)
+  call add(v:errors, 'script error: ' . v:errmsg)
+endif
+
 if empty(v:errors)
   call s:Report(['ok - all syntax assertions passed'])
   qall!