Completed the mini Clojure reader as the core of the indentation system

Some refactoring should be possible here and further optimisations. Once all optimisations I can think of have been implemented, I'll try writing an alternate Vim9 script version. (The syntax highlight group checks used in previous implementations of the indentation code was the core bottleneck, so a Vim9 script version would not have been much faster.)
clojure-vim · axvr · Apr 21, 2023 · Apr 21, 2023 · Apr 21, 2023 · Apr 21, 2023
commit ee2acc20b1b23c76a4f232e344673c61345b0704
diff --git a/indent/clojure.vim b/indent/clojure.vim
@@ -21,32 +21,36 @@ setlocal noautoindent nosmartindent nolisp
 setlocal softtabstop=2 shiftwidth=2 expandtab
 setlocal indentkeys=!,o,O
 
-" Returns true if char_idx is preceded by an odd number of backslashes.
-function! s:IsEscaped(line_str, char_idx)
-	let ln = a:line_str[: a:char_idx - 1]
+" TODO: After all optimisations create Vim9script variant of the core algorithm.
+
+" Returns "1" if position "i_char" in "line_str" is preceded by an odd number
+" of backslash characters (i.e. escaped).
+function! s:IsEscaped(line_str, i_char)
+	let ln = a:line_str[: a:i_char - 1]
 	return (strlen(ln) - strlen(trim(ln, '\', 2))) % 2
 endfunction
 
-let s:pairs = {'(': ')', '[': ']', '{': '}'}
-
-" TODO: Maybe write a Vim9script version of this?
-" Repeatedly search for tokens on the given line in reverse order building up
-" a list of tokens and their positions.  Ignores escaped tokens.
-function! s:AnalyseLine(line_num)
+" Repeatedly search for tokens on a given line (in reverse order) building up
+" a list of tokens and their positions.  Ignores escaped tokens.  Does not
+" care about strings, as that is handled by "s:InsideForm".
+function! s:TokeniseLine(line_num)
 	let tokens = []
 	let ln = getline(a:line_num)
 
 	while 1
-		" Due to legacy Vimscript being painfully slow, we literally
-		" have to move the cursor and perform searches which is
-		" ironically faster than for looping by character.
+		" We perform searches within the buffer (and move the cusor)
+		" for better performance than looping char by char in a line.
 		let token = searchpos('[()\[\]{};"]', 'bW', a:line_num)
 
+		" No more matches, exit loop.
 		if token == [0, 0] | break | endif
+
 		let t_idx = token[1] - 1
+
+		" Escaped character, ignore.
 		if s:IsEscaped(ln, t_idx) | continue | endif
-		let t_char = ln[t_idx]
 
+		let t_char = ln[t_idx]
 		if t_char ==# ';'
 			" Comment found, reset the token list for this line.
 			tokens = []
@@ -59,61 +63,79 @@ function! s:AnalyseLine(line_num)
 	return tokens
 endfunction
 
-" This should also be capable of figuring out if we're in a multi-line string
-" or regex.
-function! s:InverseRead(lnum)
-	let lnum = a:lnum - 1
+let s:pairs = {'(': ')', '[': ']', '{': '}'}
+
+" TODO: refactor this procedure and optimise.
+" This procedure is essentially a lightweight Clojure reader.
+function! s:InsideForm(lnum)
+	" Reset cursor to first column of the line we wish to indent.
+	call cursor(a:lnum, 1)
+
+	" Token list looks like this: "[[delim, [line, col]], ...]".
 	let tokens = []
+	let first_string_pos = []
+	let in_string = 0
 
+	let lnum = a:lnum - 1
 	while lnum > 0
-		call cursor(lnum + 1, 1)
-		let line_tokens = s:AnalyseLine(lnum)
+		" Reduce tokens from line "lnum" into "tokens".
+		for tk in s:TokeniseLine(lnum)
+			" Keep track of the first string delimiter we see, as
+			" we'll need it later for multi-line strings/regexps.
+			if first_string_pos == [] && tk[0] ==# '"'
+				let first_string_pos = tk[1]
+			endif
 
-		" let should_ignore = empty(a:tokens) ? 0 : (a:tokens[-1][0] ==# '"')
+			" When in string ignore other tokens.
+			if in_string && tk[0] !=# '"'
+				continue
+			else
+				let in_string = 0
+			endif
 
-		" Reduce "tokens" and "line_tokens".
-		for t in line_tokens
-			" TODO: attempt early termination.
+			" TODO: early termination?
 			if empty(tokens)
-				call add(tokens, t)
-			elseif t[0] ==# '"' && tokens[-1][0] ==# '"'
-				" TODO: track original start and ignore values
-				" inside strings.
+				call add(tokens, tk)
+			elseif tk[0] ==# '"' && tokens[-1][0] ==# '"'
 				call remove(tokens, -1)
-			elseif get(s:pairs, t[0], '') ==# tokens[-1][0]
+			elseif get(s:pairs, tk[0], '') ==# tokens[-1][0]
 				" Matching pair: drop the last item in tokens.
 				call remove(tokens, -1)
 			else
 				" No match: append to token list.
-				call add(tokens, t)
+				call add(tokens, tk)
 			endif
 		endfor
 
 		" echom 'Pass' lnum tokens
 
 		if ! empty(tokens) && has_key(s:pairs, tokens[0][0])
-			" TODO: on string match, check if string or regex.
-			" echom 'Match!' tokens[0]
 			return tokens[0]
 		endif
 
 		let lnum -= 1
 	endwhile
 
+	if ! empty(tokens) && tokens[0][0] ==# '"'
+		" Must have been in a multi-line string or regular expression
+		" as the string was never closed.
+		return ['"', first_string_pos]
+	endif
+
 	return ['^', [0, 0]]  " Default to top-level.
 endfunction
 
+" Get the value of a configuration option.
 function! s:Conf(opt, default)
 	return get(b:, a:opt, get(g:, a:opt, a:default))
 endfunction
 
+" Returns "1" when the previous operator used was "=" and is currently active.
 function! s:EqualsOperatorInEffect()
-	" Returns 1 when the previous operator used is "=" and is currently in
-	" effect (i.e. "state" includes "o").
 	return v:operator ==# '=' && state('o') ==# 'o'
 endfunction
 
-function! s:GetStringIndent(delim_pos, is_regex)
+function! s:StringIndent(delim_pos)
 	" Mimic multi-line string indentation behaviour in VS Code and Emacs.
 	let m = mode()
 	if m ==# 'i' || (m ==# 'n' && ! s:EqualsOperatorInEffect())
@@ -124,43 +146,45 @@ function! s:GetStringIndent(delim_pos, is_regex)
 		"  1: Indent in alignment with string start delimiter.
 		if     alignment == -1 | return 0
 		elseif alignment ==  1 | return a:delim_pos[1]
-		else                   | return a:delim_pos[1] - (a:is_regex ? 2 : 1)
+		else
+			let col = a:delim_pos[1]
+			let is_regex = col > 1 && getline(a:delim_pos[0])[col - 2] ==# '#'
+			return col - (is_regex ? 2 : 1)
 		endif
 	else
 		return -1  " Keep existing indent.
 	endif
 endfunction
 
-function! s:GetListIndent(delim_pos)
-	" TODO Begin analysis and apply rules!
+function! s:ListIndent(delim_pos)
 	" let lns = getline(delim_pos[0], v:lnum - 1)
-	let ln1 = getline(delim_pos[0])
-	let sym = get(split(ln1[delim_pos[1]:], '[[:space:],;()\[\]{}@\\"^~`]', 1), 0, -1)
+	let ln1 = getline(a:delim_pos[0])
+	let delim_col = a:delim_pos[1]
+	let sym = get(split(ln1[delim_col:], '[[:space:],;()\[\]{}@\\"^~`]', 1), 0, -1)
 	if sym != -1 && ! empty(sym) && match(sym, '^[0-9:]') == -1
 		" TODO: align indentation.
 		" TODO: lookup rules.
-		return delim_pos[1] + 1  " 2 space indentation
+		return delim_col + 1  " 2 space indentation
 	endif
 
 	" TODO: switch between 1 vs 2 space indentation.
-	return delim_pos[1]  " 1 space indentation
+	return delim_col  " 1 space indentation
 endfunction
 
-function! s:GetClojureIndent()
+function! s:ClojureIndent()
 	" Calculate and return indent to use based on the matching form.
-	let [formtype, coord] = s:InverseRead(v:lnum)
-	if     formtype ==# '^'  | return 0  " At top-level, no indent.
-	elseif formtype ==# '('  | return s:GetListIndent(coord)
-	elseif formtype ==# '['  | return coord[1]  " Vector
-	elseif formtype ==# '{'  | return coord[1]  " Map/set
-	elseif formtype ==# '"'  | return s:GetStringIndent(coord, 0)
-	elseif formtype ==# '#"' | return s:GetStringIndent(coord, 1)
-	else                     | return -1  " Keep existing indent.
+	let [form, pos] = s:InsideForm(v:lnum)
+	if     form ==# '^' | return 0  " At top-level, no indent.
+	elseif form ==# '(' | return s:ListIndent(pos)
+	elseif form ==# '[' | return pos[1]
+	elseif form ==# '{' | return pos[1]
+	elseif form ==# '"' | return s:StringIndent(pos)
+	else                | return -1  " Keep existing indent.
 	endif
 endfunction
 
-" TODO: lispoptions if exists.
-setlocal indentexpr=s:GetClojureIndent()
+" TODO: set lispoptions if exists.
+setlocal indentexpr=s:ClojureIndent()
 
 let &cpoptions = s:save_cpo
 unlet! s:save_cpo