test: isWellFormed (PR 774) break the lib

[cesco69]

isWellFormed don't works like the regex /[\u0000-\u001f\u0022\u005c\ud800-\udfff]/

This commit must be reverted
7f48cb6

This PR #774 just pass the test, because no test match the condition >40 chars <5000

I have added a test (serialize medium string) that enter in this condition

# Subtest: serialize short string
ok 362 - serialize short string

# Subtest: serialize medium string
not ok 363 - serialize medium string

# Subtest: serialize long string
ok 364 - serialize long string

The regex [\u0000-\u001f\u0022\u005c\ud800-\udfff]/ check for:

1. ASCII (U+0000–U+001F)
2. Char "
3. Char \
4. UTF-16 (\uD800–\uDFFF)

isWellFormed() only check case 4

These are case need escape but isWellFormed says are good!

'"'.isWellFormed()
true

'\\'.isWellFormed()
true

'\n'.isWellFormed()
true

[nigrosimone]

isWellFormed return true for string need escaping!

const str = 'abc\ndef';
str.isWellFormed()
true
JSON.stringify(str)
'"abc\\ndef"'

[gurgunday]

Seems to be true

[gurgunday]

@Uzlopak wdyt?

[Uzlopak]

@gurgunday

Well, @cesco69 ist right.

According to MDN isWellFormed just gives true if the string does not contain lone surrogates. Thats what @cesco69 means with only case 4 is covered.

So to fix this we have two options:

1. revert to the old regex,
2. use isWellFormed in combination with a shorter regex, like this pseudocode

const ASCII_ESCAPE = /[\u0000-\u001f\u0022\u005c]/u

if (str.isWellFormed() && ASCII_ESCAPE.test(str) === false) {
  // we need a bigger boat
}

We should benchmark the options.

[cesco69]

see #792