feat(vim.gsplit): gain features of vim.split

Problem:
- vim.split has more features than vim.gsplit.
- Cannot inspect the "separator" segments of vim.split or vim.gsplit.

Solution:
- Move common implementation from vim.split into vim.gsplit.
  - TODO: deprecate vim.split in favor of vim.totable(vim.gsplit())?
- Introduce `keepsep` parameter.

Related: 84f66909e4

runtime/doc/lua.txt

@@ -1649,14 +1649,25 @@ endswith({s}, {suffix})                                       *vim.endswith()*
     Return: ~
         (boolean) `true` if `suffix` is a suffix of `s`
 
-gsplit({s}, {sep}, {plain})                                     *vim.gsplit()*
+gsplit({s}, {sep}, {opts})                                      *vim.gsplit()*
     Splits a string at each instance of a separator.
 
+    Example: >lua
+
+       for s in vim.gsplit(':aa::b:', ':', {plain=true}) do
+         print(s)
+       end
+<
+
     Parameters: ~
-      • {s}      (string) String to split
+      • {s}     (string) String to split
-      • {sep}    (string) Separator or pattern
+      • {sep}   (string) Separator or pattern
-      • {plain}  (boolean|nil) If `true` use `sep` literally (passed to
+      • {opts}  (table|nil) Keyword arguments |kwargs|:
-                 string.find)
+                • keepsep: (boolean) Return segments matching `sep` instead of
+                  discarding them.
+                • plain: (boolean) Use `sep` literally (as in string.find).
+                • trimempty: (boolean) Discard empty segments at start and end
+                  of the sequence.
 
     Return: ~
         (function) Iterator over the split components
@@ -1729,7 +1740,7 @@ spairs({t})                                                     *vim.spairs()*
     See also: ~
       • Based on https://github.com/premake/premake-core/blob/master/src/base/table.lua
 
-split({s}, {sep}, {kwargs})                                      *vim.split()*
+split({s}, {sep}, {opts})                                        *vim.split()*
     Splits a string at each instance of a separator.
 
     Examples: >lua
@@ -1738,16 +1749,14 @@ split({s}, {sep}, {kwargs})                                      *vim.split()*
       split("axaby", "ab?")                   --> {'','x','y'}
       split("x*yz*o", "*", {plain=true})      --> {'x','yz','o'}
       split("|x|y|z|", "|", {trimempty=true}) --> {'x', 'y', 'z'}
+      split("|x|y|z|", "|", {keepsep=true})   --> {'|', 'x', '|', 'y', '|', 'z', '|'}
 <
 
     Parameters: ~
-      • {s}       (string) String to split
+      • {s}     (string) String to split
-      • {sep}     (string) Separator or pattern
+      • {sep}   (string) Separator or pattern
-      • {kwargs}  (table|nil) Keyword arguments:
+      • {opts}  (table|nil) Keyword arguments |kwargs| accepted by
-                  • plain: (boolean) If `true` use `sep` literally (passed to
+                |vim.gsplit()|
-                    string.find)
-                  • trimempty: (boolean) If `true` remove empty items from the
-                    front and back of the list
 
     Return: ~
         string[] List of split components

runtime/doc/news.txt

@@ -132,6 +132,8 @@ The following new APIs or features were added.
 • |vim.fs.dir()| now has a `opts` argument with a depth field to allow
   recursively searching a directory tree.
 
+• |vim.gsplit()| supports all features of |vim.split()|.
+
 • |vim.secure.read()| reads a file and prompts the user if it should be
   trusted and, if so, returns the file's contents.
 

runtime/lua/vim/shared.lua

@@ -59,6 +59,13 @@ end)()
 
 --- Splits a string at each instance of a separator.
 ---
+--- Example:
+---   <pre>lua
+---   for s in vim.gsplit(':aa::b:', ':', {plain=true}) do
+---     print(s)
+---   end
+---   </pre>
+---
 ---@see |vim.split()|
 ---@see |luaref-patterns|
 ---@see https://www.lua.org/pil/20.2.html
@@ -66,17 +73,40 @@ end)()
 ---
 ---@param s string String to split
 ---@param sep string Separator or pattern
----@param plain (boolean|nil) If `true` use `sep` literally (passed to string.find)
+---@param opts (table|nil) Keyword arguments |kwargs|:
----@return fun():string (function) Iterator over the split components
+---       - keepsep: (boolean) Include segments matching `sep` instead of discarding them.
-function vim.gsplit(s, sep, plain)
+---       - plain: (boolean) Use `sep` literally (as in string.find).
-  vim.validate({ s = { s, 's' }, sep = { sep, 's' }, plain = { plain, 'b', true } })
+---       - trimempty: (boolean) Discard empty segments at start and end of the sequence.
+---@return fun():string|nil (function) Iterator over the split components
+function vim.gsplit(s, sep, opts)
+  local plain
+  local trimempty = false
+  local keepsep = false
+  if type(opts) == 'boolean' then
+    plain = opts -- For backwards compatibility.
+  else
+    vim.validate({ s = { s, 's' }, sep = { sep, 's' }, opts = { opts, 't', true } })
+    opts = opts or {}
+    plain, trimempty, keepsep = opts.plain, opts.trimempty, opts.keepsep
+    assert(not trimempty or not keepsep, 'keepsep+trimempty not supported')
+  end
 
   local start = 1
   local done = false
+  local sepseg = nil -- Last matched `sep` segment.
+  local sepesc = plain and vim.pesc(sep) or sep
+
+  -- For `trimempty`:
+  local empty_start = true -- Only empty segments seen so far.
+  local empty_segs = 0 -- Empty segments found between non-empty segments.
+  local nonemptyseg = nil
 
   local function _pass(i, j, ...)
     if i then
       assert(j + 1 > start, 'Infinite loop detected')
+      if keepsep then
+        sepseg = s:match(sepesc, start)
+      end
       local seg = s:sub(start, i - 1)
       start = j + 1
       return seg, ...
@@ -87,16 +117,48 @@ function vim.gsplit(s, sep, plain)
   end
 
   return function()
-    if done or (s == '' and sep == '') then
+    if trimempty and empty_segs > 0 then
-      return
+      -- trimempty: Pop the collected empty segments.
-    end
+      empty_segs = empty_segs - 1
-    if sep == '' then
+      return ''
+    elseif trimempty and nonemptyseg then
+      local seg = nonemptyseg
+      nonemptyseg = nil
+      return seg
+    elseif keepsep and sepseg then
+      local seg = sepseg
+      sepseg = nil
+      return seg
+    elseif done or (s == '' and sep == '') then
+      return nil
+    elseif sep == '' then
       if start == #s then
         done = true
       end
       return _pass(start + 1, start)
     end
-    return _pass(s:find(sep, start, plain))
+
+    local seg = _pass(s:find(sep, start, plain))
+
+    -- Trim empty segments from start/end.
+    if trimempty and seg == '' then
+      while not done and seg == '' do
+        empty_segs = empty_segs + 1
+        seg = _pass(s:find(sep, start, plain))
+      end
+      if done and seg == '' then
+        return nil
+      elseif empty_start then
+        empty_start = false
+        empty_segs = 0
+        return seg
+      end
+      nonemptyseg = seg ~= '' and seg or nil
+      seg = ''
+      empty_segs = empty_segs - 1
+    end
+
+    return seg
   end
 end
 
@@ -108,51 +170,21 @@ end
 ---  split("axaby", "ab?")                   --> {'','x','y'}
 ---  split("x*yz*o", "*", {plain=true})      --> {'x','yz','o'}
 ---  split("|x|y|z|", "|", {trimempty=true}) --> {'x', 'y', 'z'}
+---  split("|x|y|z|", "|", {keepsep=true})   --> {'|', 'x', '|', 'y', '|', 'z', '|'}
 --- </pre>
 ---
 ---@see |vim.gsplit()|
 ---
 ---@param s string String to split
 ---@param sep string Separator or pattern
----@param kwargs (table|nil) Keyword arguments:
+---@param opts (table|nil) Keyword arguments |kwargs| accepted by |vim.gsplit()|
----       - plain: (boolean) If `true` use `sep` literally (passed to string.find)
----       - trimempty: (boolean) If `true` remove empty items from the front
----         and back of the list
 ---@return string[] List of split components
-function vim.split(s, sep, kwargs)
+function vim.split(s, sep, opts)
-  local plain
+  -- TODO(justinmk): deprecate vim.split in favor of vim.totable(vim.gsplit())
-  local trimempty = false
-  if type(kwargs) == 'boolean' then
-    -- Support old signature for backward compatibility
-    plain = kwargs
-  else
-    vim.validate({ kwargs = { kwargs, 't', true } })
-    kwargs = kwargs or {}
-    plain = kwargs.plain
-    trimempty = kwargs.trimempty
-  end
-
   local t = {}
-  local skip = trimempty
+  for c in vim.gsplit(s, sep, opts) do
-  for c in vim.gsplit(s, sep, plain) do
+    table.insert(t, c)
-    if c ~= '' then
-      skip = false
-    end
-
-    if not skip then
-      table.insert(t, c)
-    end
   end
-
-  if trimempty then
-    for i = #t, 1, -1 do
-      if t[i] ~= '' then
-        break
-      end
-      table.remove(t, i)
-    end
-  end
-
   return t
 end
 

test/functional/lua/vim_spec.lua

@@ -292,51 +292,62 @@ describe('lua stdlib', function()
     ]]}
   end)
 
-  it("vim.split", function()
+  it('vim.gsplit, vim.split', function()
-    local split = function(str, sep, kwargs)
-      return exec_lua('return vim.split(...)', str, sep, kwargs)
-    end
-
     local tests = {
-      { "a,b", ",", false, false, { 'a', 'b' } },
+      { 'a,b',             ',',     false, false, { 'a', 'b' } },
-      { ":aa::bb:", ":", false, false, { '', 'aa', '', 'bb', '' } },
+      { ':aa::::bb:',      ':',     false, false, { '', 'aa', '', '', '', 'bb', '' } },
-      { ":aa::bb:", ":", false, true, { 'aa', '', 'bb' } },
+      { ':aa::::bb:',      ':',     false, true,  { 'aa', '', '', '', 'bb' } },
-      { "::ee::ff:", ":", false, false, { '', '', 'ee', '', 'ff', '' } },
+      { ':aa::bb:',        ':',     false, true,  { 'aa', '', 'bb' } },
-      { "::ee::ff:", ":", false, true, { 'ee', '', 'ff' } },
+      { '/a/b:/b/\n',      '[:\n]', false, true,  { '/a/b', '/b/' } },
-      { "ab", ".", false, false, { '', '', '' } },
+      { '::ee::ff:',       ':',     false, false, { '', '', 'ee', '', 'ff', '' } },
-      { "a1b2c", "[0-9]", false, false, { 'a', 'b', 'c' } },
+      { '::ee::ff::',      ':',     false, true,  { 'ee', '', 'ff' } },
-      { "xy", "", false, false, { 'x', 'y' } },
+      { 'ab',              '.',     false, false, { '', '', '' } },
-      { "here be dragons", " ", false, false, { "here", "be", "dragons"} },
+      { 'a1b2c',           '[0-9]', false, false, { 'a', 'b', 'c' } },
-      { "axaby", "ab?", false, false, { '', 'x', 'y' } },
+      { 'xy',              '',      false, false, { 'x', 'y' } },
-      { "f v2v v3v w2w ", "([vw])2%1", false, false, { 'f ', ' v3v ', ' ' } },
+      { 'here be dragons', ' ',     false, false, { 'here', 'be', 'dragons'} },
-      { "", "", false, false, {} },
+      { 'axaby',           'ab?',   false, false, { '', 'x', 'y' } },
-      { "", "a", false, false, { '' } },
+      { 'f v2v v3v w2w ',  '([vw])2%1', false, false, { 'f ', ' v3v ', ' ' } },
-      { "x*yz*oo*l", "*", true, false, { 'x', 'yz', 'oo', 'l' } },
+      { '',                '',      false, false, {} },
+      { '',                '',      false, true,  {} },
+      { '\n',              '[:\n]', false, true,  {} },
+      { '',                'a',     false, false, { '' } },
+      { 'x*yz*oo*l',       '*',     true,  false, { 'x', 'yz', 'oo', 'l' } },
     }
 
     for _, t in ipairs(tests) do
-      eq(t[5], split(t[1], t[2], {plain=t[3], trimempty=t[4]}))
+      eq(t[5], vim.split(t[1], t[2], {plain=t[3], trimempty=t[4]}))
     end
 
     -- Test old signature
-    eq({'x', 'yz', 'oo', 'l'}, split("x*yz*oo*l", "*", true))
+    eq({'x', 'yz', 'oo', 'l'}, vim.split("x*yz*oo*l", "*", true))
 
     local loops = {
       { "abc", ".-" },
     }
 
     for _, t in ipairs(loops) do
-      matches("Infinite loop detected", pcall_err(split, t[1], t[2]))
+      matches("Infinite loop detected", pcall_err(vim.split, t[1], t[2]))
     end
 
+    -- `keepsep`
+    eq({ '', '.', '', '.', 'aa', '.', 'bb', '.', 'cc', '.', 'dd', '.', 'ee', '.', '', },
+      vim.split('..aa.bb.cc.dd.ee.', '%.', {keepsep=true}))
+    eq({ '..aa', '1', '.bb', '2', '', '2', '.cc.', '9', '', },
+      vim.split('..aa1.bb22.cc.9', '%d', {keepsep=true}))
+    eq({ '..aa', '1', '.bb', '22', '.cc.', '9', '', },
+      vim.split('..aa1.bb22.cc.9', '%d+', {keepsep=true}))
+
     -- Validates args.
-    eq(true, pcall(split, 'string', 'string'))
+    eq(true, pcall(vim.split, 'string', 'string'))
     matches('s: expected string, got number',
-      pcall_err(split, 1, 'string'))
+      pcall_err(vim.split, 1, 'string'))
     matches('sep: expected string, got number',
-      pcall_err(split, 'string', 1))
+      pcall_err(vim.split, 'string', 1))
-    matches('kwargs: expected table, got number',
+    matches('opts: expected table, got number',
-      pcall_err(split, 'string', 'string', 1))
+      pcall_err(vim.split, 'string', 'string', 1))
+    -- Not supported (yet).
+    matches('keepsep%+trimempty not supported',
+      pcall_err(vim.split, 'foo bar', ' ', {keepsep=true, trimempty=true}))
   end)
 
   it('vim.trim', function()