Re: [GSoC][PATCH v3] Add a diff driver for JavaScript languages.

["Ævar Arnfjörð Bjarmason" <avarab@gmail.com>]

On Sun, Apr 03 2022, xing zhi jiang wrote:

Aside from what Johannes Sixt mentioned:

> +PATTERNS("javascript",
> +	 /* don't match the expression may contain parenthesis, because it is not a function declaration */
> +	 "!^[ \t]*(if|do|while|for|with|switch|catch|import|return)\n"
> +	 /* don't match statement */
> +	 "!;\n"
> +	 /* match normal function or named export for function in ECMA2015 */
> +	 "^((export[\t ]+)?(async[\t ]+)?function[\t ]*[\t *]*[$_[:alpha:]][$_[:alnum:]]*[\t ]*\\(.*)\n"
> +	 /* match JavaScript variable declaration with a lambda expression at top level */
> +	 "^((const|let|var)[\t ]*[$_[:alpha:]][$_[:alnum:]]*[\t ]*=[\t ]*"
> +		"(\\(.*\\)|[$_[:alpha:]][$_[:alnum:]]*)[\t ]*=>[\t ]*\\{?)\n"
> +	 /* match object's property assignment by anonymous function and CommonJS exports for named function */
> +	 "^((module\\.)?[$_[:alpha:]][$_[:alnum:]]*\\.[$_[:alpha:]][$_[:alnum:]]*[\t ]*=[\t ]*(async[\t ]+)?(\\(.*\\)|[$_[:alpha:]][$_[:alnum:]]*)[\t ]*=>.*)\n"
> +	 /* match assign function to LHS with explicit function keyword */
> +	 "^(.*=[\t ]*function[\t ]*([$_[:alnum:]]+[\t ]*)?\\(.*)\n"
> +	 /* popular unit testing framework test case pattern. Most of framework pattern is match by regex for "function in class" */
> +	 "^[\t ]*(QUnit.test\\(.*)\n"
> +	 /* don't match the function in class or in object literal, which has more than one ident level */
> +	 "!^(\t{2,}|[ ]{5,})\n"
> +	 /* match normal function in object literal */
> +	 "^[\t ]*([$_[:alpha:]][$_[:alnum:]]*[\t ]*:[\t ]*function.*)\n"
> +	 /* don't match chained method call */
> +	 "!^[\t ]*[$_[:alpha:]][$_[:alnum:]][\t ]*\\(.*\\)\\.\n"
> +	 /* match function in class and ES5 method shorthand */
> +	 "^[\t ]*((static[\t ]+)?((async|get|set)[\t ]+)?[$_[:alpha:]][$_[:alnum:]]*[\t ]*\\(.*)",
> +	 /* word regex */
> +	 /* hexIntegerLiteral, octalIntegerLiteral, binaryIntegerLiteral, and its big version */
> +	 "0[xXoObB][_0-9a-fA-F]+n?"
> +	 /* DecimalLiteral and its big version*/
> +	 "|[0-9][_0-9]*(\\.[0-9][_0-9]*|n)?([eE][+-]?[_0-9]+)?"
> +	 "|\\.[0-9][_0-9]*([eE][+-]?[_0-9]+)?"
> +	 /* punctuations */
> +	 "|\\.{3}|<=|>=|==|!=|={3}|!==|\\*{2}|\\+{2}|--|<<|>>"
> +	 "|>>>|&&|\\|{2}|\\?{2}|\\+=|-=|\\*=|%=|\\*{2}="
> +	 "|<<=|>>=|>>>=|&=|\\|=|\\^=|&&=|\\|{2}=|\\?{2}=|=>"
> +	 /* identifiers */
> +	 "|[$_[:alpha:]][$_[:alnum:]]*"),
>  PATTERNS("markdown",
>  	 "^ {0,3}#{1,6}[ \t].*",
>  	 /* -- */<

While we don't use helper macros for these currently there's no reason
we can't, I thin the above might be more readable with e.g.:

	#define JS_AA "[$_[:alpha:]][$_[:alnum:]]"

Which would make this:
	
	+PATTERNS("javascript",
	+	 /* don't match the expression may contain parenthesis, because it is not a function declaration */
	+	 "!^[ \t]*(if|do|while|for|with|switch|catch|import|return)\n"
	+	 /* don't match statement */
	+	 "!;\n"
	+	 /* match normal function or named export for function in ECMA2015 */
	+	 "^((export[\t ]+)?(async[\t ]+)?function[\t ]*[\t *]*" JS_AA "*[\t ]*\\(.*)\n"
	+	 /* match JavaScript variable declaration with a lambda expression at top level */
	+	 "^((const|let|var)[\t ]*" JS_AA "*[\t ]*=[\t ]*"
	+		"(\\(.*\\)|" JS_AA "*)[\t ]*=>[\t ]*\\{?)\n"
	+	 /* match object's property assignment by anonymous function and CommonJS exports for named function */
	+	 "^((module\\.)?" JS_AA "*\\." JS_AA "*[\t ]*=[\t ]*(async[\t ]+)?(\\(.*\\)|" JS_AA "*)[\t ]*=>.*)\n"
	+	 /* match assign function to LHS with explicit function keyword */
	+	 "^(.*=[\t ]*function[\t ]*([$_[:alnum:]]+[\t ]*)?\\(.*)\n"
	+	 /* popular unit testing framework test case pattern. Most of framework pattern is match by regex for "function in class" */

Wry try to stick to wrapping at 80 characters, so some of these comments
should really be wrapped (see CodingGuidelines for the multi-line
comment style we use).

	+	 "^[\t ]*(QUnit.test\\(.*)\n"
	+	 /* don't match the function in class or in object literal, which has more than one ident level */
	+	 "!^(\t{2,}|[ ]{5,})\n"
	+	 /* match normal function in object literal */
	+	 "^[\t ]*(" JS_AA "*[\t ]*:[\t ]*function.*)\n"
	+	 /* don't match chained method call */
	+	 "!^[\t ]*" JS_AA "[\t ]*\\(.*\\)\\.\n"
	+	 /* match function in class and ES5 method shorthand */
	+	 "^[\t ]*((static[\t ]+)?((async|get|set)[\t ]+)?" JS_AA "*[\t ]*\\(.*)",
	+	 /* word regex */
	+	 /* hexIntegerLiteral, octalIntegerLiteral, binaryIntegerLiteral, and its big version */
	+	 "0[xXoObB][_0-9a-fA-F]+n?"
	+	 /* DecimalLiteral and its big version*/
	+	 "|[0-9][_0-9]*(\\.[0-9][_0-9]*|n)?([eE][+-]?[_0-9]+)?"
	+	 "|\\.[0-9][_0-9]*([eE][+-]?[_0-9]+)?"
	+	 /* punctuations */
	+	 "|\\.{3}|<=|>=|==|!=|={3}|!==|\\*{2}|\\+{2}|--|<<|>>"
	+	 "|>>>|&&|\\|{2}|\\?{2}|\\+=|-=|\\*=|%=|\\*{2}="
	+	 "|<<=|>>=|>>>=|&=|\\|=|\\^=|&&=|\\|{2}=|\\?{2}=|=>"
	+	 /* identifiers */
	+	 "|" JS_AA "*"),
	
Just a thought, I wonder how much line-noisy we could make this thing in
general if we defined some common patterns with such helpers.

Anyway, insted of :alnum:and :alpha: don't you really mean [a-zA-Z0-9]
and [a-zA-Z]. I.e. do you *really* want to have this different depending
on the user's locale?

I haven't tested, but see the LC_CTYPE in gettext.c, so I'm fairly sure
that'll happen...