feat: decode percent-encoded path in getPath

[usualoma]

#2672

This PR contains #2688 + #2688 (comment)

Specification Changes

There are no major release-level changes, but there are some minor changes as follows

Symbols and multibyte characters can be included in the static string of the routing definition

When you define a routing as follows, it will now match, whereas before it did not match because the percent encoding for static strings was not taken into account when routing.

app.get('/🔥', (c) => c.text(`I'm hono`)

http://example.com/%F0%9F%94%A5

Requests made with a URL that are encoded more than necessary are now matched with a decoded URL

Suppose we have a routing definition like this,

app.get('/static', (c) => c.text('static)

Even if unreserved characters are extra encoded, they are now decoded and matched.

http://example.com/%73tatic // %73 is 's'

Regular expressions for path parameters are now tested against decoded characters

We do not think that this will break any applications, but applications that use regular expressions in path parameters should check when updating to see if they are affected.

Suppose we have a routing definition like this,

app.get('/about/:product{🔥|Hono}', (c) => c.text(`Hono is a framework`)

The following URLs match

http://example.com/about/%F0%9F%94%A5

The author should do the following, if applicable

• Add tests
• Run tests
• bun denoify to generate files for Deno
• bun run format:fix && bun run lint:fix to format the code

[szmarczak]

From my benchmarks it seems like str.includes('%25') ? str.replaceAll('%25', '%2525') : str is faster than calling .replace with regex. Can you verify?

[usualoma]

Thank you. You are right!

It's faster to check in .includes first.
s/replace/replaceAll/ didn't change in my benchmark results, so I'll leave replace(//g) as it is, as it is generally used elsewhere in Hono.

Applied in 0eeaa7e

[yusukebe]

@usualoma

Thanks! I'll include this in the next minor release and merge it into the next branch later.