[LNT] Python 3 support: fix convert to JSON

Plistlib and json modules have conflicting requirements for their dump
methods: Plistlib expects its file handle to be a writable binary file
object while json expects it to be a writable text file object.

This commit solves this issue by dumping the JSON into a temporary
string which is encoded to UTF-8 (JSON format RFC 7159 requires UTF-8,
UTF-16 or UTF-32 encoding and UTF-8 is the recommended default) before
being output. This allows to keep delegating file opening to Click and
avoid a refactoring to abstract the opening mode between supported
formats.

Reviewers: cmatthews, hubert.reinterpretcast, kristof.beyls

Reviewed By: hubert.reinterpretcast

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D68223

llvm-svn: 373615

Author: Thomas Preud'homme

lnt/formats/JSONFormat.py

@@ -19,9 +19,17 @@ def _load_format(path_or_file):
     return json.load(path_or_file)
 
 
+def _dump_format(obj, fp):
+    # The json module produces str objects but fp is opened in binary mode
+    # (since Plistlib only dump to binary mode files) so we first dump into
+    # a string a convert to UTF-8 before outputing.
+    json_str = json.dumps(obj)
+    fp.write(json_str.encode())
+
+
 format = {
     'name': 'json',
     'predicate': _matches_format,
     'read': _load_format,
-    'write': json.dump,
+    'write': _dump_format,
 }