This patch adds a command line option "--no-escape" that allows json.tool to display non-ASCII characters. e.g.: $ echo '"測試"' | python -m json.tool "\u6e2c\u8a66" $ echo '"測試"' | python -m json.tool --no-escape "測試"

Maybe name it --no-insure-ascii? Or --insure-ascii=no/yes. (or true/false).

If the arguments should be aligned with those in dump/load, then maybe "--no-ensure-ascii" is an option?

Sorry, yes, that's what I meant. I think it will make it easier to understand and remember if the option uses the same terminology as the function.

Use "--no-ensure-ascii" instead.

The patch needs tests and documentation. > + parser.add_argument('--no-ensure-ascii', action='store_true', default=False, I'd go with ``action='store_false', default=True``.

> The patch needs tests and documentation. Ok, I'll update it later. >> + parser.add_argument('--no-ensure-ascii', action='store_true', default=False, >I'd go with ``action='store_false', default=True``. If I'm not misreading your comment, this will change the original behavior, right? (because options.no_ensure_ascii will be set to True by default)

Added doc and test.

Test fails on non-utf8 locale. $ LC_ALL=en_US ./python -m test.regrtest -v -m test_no_ensure_ascii_flag test_json ... ====================================================================== FAIL: test_no_ensure_ascii_flag (test.test_json.test_tool.TestTool) ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/serhiy/py/cpython/Lib/test/test_json/test_tool.py", line 115, in test_no_ensure_ascii_flag self.assertEqual(out.splitlines(), b'"\\u6e2c\\u8a66"\n'.splitlines()) AssertionError: Lists differ: [b'"\\u00e6\\u00b8\\u00ac\\u00e8\\u00a9\\u00a6"'] != [b'"\\u6e2c\\u8a66"'] First differing element 0: b'"\\u00e6\\u00b8\\u00ac\\u00e8\\u00a9\\u00a6"' b'"\\u6e2c\\u8a66"' - [b'"\\u00e6\\u00b8\\u00ac\\u00e8\\u00a9\\u00a6"'] + [b'"\\u6e2c\\u8a66"'] ----------------------------------------------------------------------

1. Replaced non-ASCII literals to \uXXXX 2. Removed failed assertion Test passed with LC_ALL=en_US and LC_ALL=en_US.UTF-8 . I've tried to use locale.getdefaultlocale(), but seems the output string will vary in different locales.

> If I'm not misreading your comment, this will change the original behavior, right? Assuming you also change ensure_ascii = not options.no_ensure_ascii to ensure_ascii = options.no_ensure_ascii no, it won't change the original behavior. That way you won't need to invert the value of ``options.no_ensure_ascii`` in line 37.

The last change just sweeps a problem under a rug. For now json.tool never fails with valid data. But with the --no-ensure-ascii option it can fail when output a string not encodable with the locale encoding. All can work with common cases on common UTF-8 environment, but unexpectedly fail on nonstandard environment. It would be better to output encodable characters as is and represent unencodable characters with \uXXXX encoding.

 New changeset efefe25443c56988841ab96cdac01352123ba268 by Inada Naoki (wim glenn) in branch 'master': bpo-27413: json.tool: Add --no-ensure-ascii option. (GH-17472) https://github.com/python/cpython/commit/efefe25443c56988841ab96cdac01352123ba268