ES6 tagged templates practical usability

问题:

I understand the syntax of ES6 tagged templates. What I don't see is the practical usability. When is it better than passing an object parameter, like the settings in jQuery's AJAX? $.ajax('url', { /*this guy here*/ })

Right now I only see the tricky syntax but I don't see why I would need/use it. I also found that the TypeScript team chose to implement it (in 1.5) before other important features. What is the concept behind tagged string templates?

回答1:

You can use tagged templates to build APIs that are more expressive than regular function calls.

For example, I'm working on a proof-of-concept library for SQL queries on JS arrays:

let admins = sql`SELECT name, id FROM ${users} 
                 WHERE ${user => user.roles.indexOf('admin') >= 0}`

Notice it has nothing to do with String interpolation; it uses tagged templates for readability. It would be hard to construct something that reads as intuitively with plain function calls - I guess you'd have something like this:

let admins = sql("SELECT name, id FROM $users WHERE $filter",
  { $users: users, $filter: (user) => user.roles.contains('admin') })

This example is just a fun side project, but I think it shows some of the benefits of tagged templates.

Another example, maybe more obvious, is i18n - a tagged template could insert locale-sensitive versions of your input.

回答2:

See Sitepoint's explanation:

The final stage of template strings specification is about adding a custom function before the string itself to create a tagged template string.

...

For instance, here is a piece of code to block strings that try to inject custom DOM elements:

var items = [];
items.push("banana");
items.push("tomato");
items.push("light saber");
var total = "Trying to hijack your site <BR>";
var myTagFunction = function (strings,...values) {
  var output = "";
  for (var index = 0; index < values.length; index++) {
    var valueString = values[index].toString();

    if (valueString.indexOf(">") !== -1) {
      // Far more complex tests can be implemented here :)
      return "String analyzed and refused!";
    }

    output += strings[index] + values[index];
  }

  output += strings[index]
  return output;
}

result.innerHTML = myTagFunction `You have ${items.length} item(s) in your basket for a total of $${total}`;

Tagged template strings can used for a lot of things like security, localization, creating your own domain specific language, etc.

回答3:

They're useful because the function can (almost) completely define the meaning of the text inside it (almost = other than placeholders). I like to use the example of Steven Levithan's XRegExp library. It's awkward to use regular expressions defined as strings, because you have to double-escape things: Once for the string literal, and once for regex. This is one of the reasons we have regular expression literals in JavaScript.

For instance, suppose I'm doing maintenance on a site and I find this:

var isSingleUnicodeWord = /^\w+$/;

...which is meant to check if a string contains only "letters." Two problems: A) There are thousands of "word" characters across the realm of human language that \w doesn't recognize, because its definition is English-centric; and B) It includes _, which many (including the Unicode consortium) would argue is not a "letter."

Suppose in my work I've introduced XRegExp to the codebase. Since I know it supports \pL (\p for Unicode categories, and L for "letter"), I might quickly swap this in:

var isSingleUnicodeWord = XRegExp("^\pL+$"); // WRONG

Then I wonder why it didn't work, *facepalm*, and go back and escape that backslash, since it's being consumed by the string literal:

var isSingleUnicodeWord = XRegExp("^\\pL+$");
// ---------------------------------^

What a pain. Suppose I could write the actual regular expression without worrying about double-escaping?

I can: With a tagged template function. I can put this in my standard lib:

function xrex(strings, ...values) {
    const raw = strings.raw;
    let result = "";
    for (let i = 0; i < raw.length; ++i) {
        result += raw[i];
        if (i < values.length) { // `values` always has one fewer entry
            result += values[i];
        }
    }
    return XRegExp(result);
}

Or alternately, this is a valid use case for reduce, and we can use destructuring in the argument list:

function xrex({raw}, ...values) {
    return XRegExp(
        raw.reduce(
            (acc, str, index) => acc + str + (index < values.length ? values[index] : ""),
            ""
        )
    );
}

And then I can happily write:

const isSingleUnicodeWord = xrex`^\pL+$`;

Example:

// My tag function (defined once, then reused)
function xrex({raw}, ...values) {
    const result = raw.reduce(
        (acc, str, index) => acc + str + (index < values.length ? values[index] : ""),
        ""
    );
    console.log("Creating with:", result);
    return XRegExp(result);
}

// Using it, with a couple of substitutions to prove to myself they work
let category = "L";                // L: Letter
let maybeEol = "$";
let isSingleUnicodeWord = xrex`^\p${category}+${maybeEol}`;
function test(str) {
    console.log(str + ": " + isSingleUnicodeWord.test(str));
}
test("Русский");  // true
test("日本語");    // true
test("العربية");  // true
test("foo bar");  // false
test("$£");       // false

<script src="https://cdnjs.cloudflare.com/ajax/libs/xregexp/3.2.0/xregexp-all.min.js"></script>

The only thing I have to remember now is that ${...} is special because it's a placeholder. In this specific case, it's not a problem, I'm unlikely to want to apply a quantifier to the end-of-input assertion, but that's a coincidence...