PHP：什么时候应该将未转义的UTF-8保存到json文件中？

Question

Is there any benefit of saving UTF-8 characters unescaped in a json file if one only access them through PHP?

Here is what I tested:

fwrite(fopen('fileA.json','w'), json_encode('аккредитовать'));  

then the content of fileA.json is given by

"\u0413\u043b\u0430\u0432\u043d\u0430\u044f"

However, when I store it with

fwrite(fopen('fileB.json','w'), json_encode('аккредитовать', JSON_UNESCAPED_UNICODE));

the content of fileB.json is given by

"аккредитовать"

To my surprise each of the following calls

echo json_decode(file_get_contents('fileA.json'));
echo json_decode(file_get_contents('fileB.json'));
echo json_decode(file_get_contents('fileA.json')), false, 512, JSON_UNESCAPED_UNICODE);
echo json_decode(file_get_contents('fileB.json')), false, 512, JSON_UNESCAPED_UNICODE);

gives the same output:

'аккредитовать'

So as a result I would conclude that I only need to save UTF-8 chars in a json file if I want to open and read the json file directly with an editor. If I only plan to show/save the content of the json file with php then I don't need save the content unescaped and I can use

fwrite(fopen('fileA.json','w'), json_encode('аккредитовать'));  
echo json_decode(file_get_contents('fileA.json'));`

Is that correct, or did I miss anything important?

Answer 1

With JSON_UNESCAPED_UNICODE the JSON is now:

1. more human readable
2. not ASCII-safe

That's the only tradeoff you're making. Once you have non-ASCII characters in your JSON, you need to ensure the JSON is handled in a binary-safe manner; e.g. you cannot simply send it over a channel that expects only ASCII data, or you need to care about the specific encoding if a channel is encoding aware (e.g. storing it in a database). None of this is of any concern when simply writing the data to a file and then reading it again, as long as the reader is treating the encoding correctly (which PHP is doing here, since it doesn't care about the encoding).

The JSON format itself doesn't care either way, "а" and "\u0413" represent the exact same character.

It should be noted that escaped Unicode takes up more storage than UTF-8 encoded text (6-12 bytes vs. 2-4 bytes). But that hardly matters in the majority of cases.

Note also: JSON_UNESCAPED_UNICODE is not a valid flag for json_decode; it's simply superfluous there.