Quoting syntax

[yshavit]

My current thoughts:

• selectors are delimited by |
• bareword strings have to start with an alpha, and go until a context-specific ending character.
  • In most contexts, that's the selector delimiter |. In links or images, it's ] for the display portion and ) for the URL portion.
  • bareword strings don't support escapes; every character except newline is literal. newlines become spaces. bareword strings are case-insensitive. If none of those restrictions work for you, use a quoted string.
• quoted strings come in two forms, both of which are case-sensitive
  • standard quoted strings: these are surrounded by " (maybe ', too?), and support standard escapes as well as a tbd unicode escape sequences. The quote char can be escaped.
  • custom quoted strings: these start with {"<custom>"} where <custom> is any string that does not contain a " or a \. The string then ends with that <custom>. Standard escapes are supported; you cannot escape the custom ender. For example: {"%%"}my cool string which can have "quotes".%%, but you can't do {"%%"}my cool \%% string. Just use a different ender! The custom string may not be empty.
• any of these strings support ^ and $ as anchors
• regular expressions are standard, except you can also have one that uses custom quoting: {/<custom>/}, where <custom> can't include / or \. This is particularly useful for URL regexes: {/%%/}^https://example.com/%%.
• empty matchers and * are both equivalent to "any"

Example:

mdq '# foo | [](^https://example.com) | ![{"#"}my "quote"#]({/%%/}^http://my-regex\.com?%%')

• Look for a header with "foo" in it
  • and within that, a link to https://example.com/.* (string are always substring search, but this example has a ^ anchor)
  • and within that link text, an image with an alt text that includes my "quote" and a URL to http://my-regex.co or http://my-regex.com. As with strings, regexes are always substring but support anchors.

[yshavit]

The custom string may not be empty.

Maybe I should allow that, so that people can add modifiers. For example:

{""i}my case-insensitive matcher

[yshavit]

bareword strings have to start with an alpha, and go until a context-specific ending character.

Maybe add "or a dollar sign", so that people can add anchors?

Or maybe anchors are only available on quoted strings.

[yshavit]

Done in #91

[yshavit]

As of b8f346c, the following are implemented:

• ✅ selectors delimited by |
• ✅ bareword strings; start with alpha, no escaping, ended by end-delimiter or $; case-insensitive
• ✅ standard quoted strings, using either " or ', with escaping; case-sensitive
  • ❌ custom quoted strings ({"EOF"}foo barEOF)
• ✅ ^ and $ as anchors; you can't have just one anchor (e.g. ^ by itself), though you can have both anchors (^$) to signify an empty string
• ✅ regular expressions
  • ❌ regular expressions with custom quoting ({/EOF/}foo #\dEOF)
• ✅ empty matchers and * are both equivalent to "any"

[yshavit]

Modifiers: Do I want to require the opening token to also close it:

• ❔ [{"#"}my "quote#]()
• ❔ [{"#}my "quote#]()

I think I do want it -- it just looks lopsided without it.

I considered:

• ❔ [#"my "quote#]()
  • but the # can be potentially ambiguous, at least visually
  • also it's not at all clear that #" means "#-delimited quote"

Another option:

• ❔ [{"#}#my "quote#]()
• it adds duplication (you have to write the delimiter an extra time, at the start)
• but in return, it's much more visually clear -- the actual string (my "quote) is clearly and obviously bookended by the delimiter
• You could also add optional whitespace between the curlies and the opener

I think I like that one best. Here's what it could look like for regexes:

{/@} @^https?://example\.com/foo/@

or small variant:

{/@/} @^https?://example\.com/foo/@

[{"#"}#my "quote#]()

I think that may be the winner so far.

For regexes, I should make it clear that the end-delimiter takes precedence over any special chars. For example:

{/$/} $^https?://example\.com/foo/$$

This does not compile. The first $ after foo/ ends the regex, and the second one is just a floating dollar sign (which isn't allowed). Since you can't escape custom delimiters, the only option (which is a good one!) is to just use a different custom delimiter.

Maybe I should prohibit $ and ^ altogether from delimiters, since it can be confusing in quoted forms:

{"$"} $foo$$

That's the string foo anchored to an end char. Actually idk, maybe that's not so bad.