CLI option --filter: validate multiple instances from stdin.

Reads JSON instances one to a line from stdin, outputs those
that validate in compact form (one to a line).  Print errors
to stderr for those that do not validate.

Author: jkleint

jsonschema/cli.py

@@ -21,14 +21,21 @@ def _json_file(path):
 parser = argparse.ArgumentParser(
     description="JSON Schema Validation CLI",
 )
-parser.add_argument(
+input_group = parser.add_mutually_exclusive_group()
+input_group.add_argument(
     "-i", "--instance",
     action="append",
     dest="instances",
     type=_json_file,
     help="a path to a JSON instance to validate "
          "(may be specified multiple times)",
 )
+input_group.add_argument(
+    "-f", "--filter",
+    action="store_true",
+    help="read instances one to a line from stdin, output those that validate "
+         "in compact form and log to stderr those that do not."
+)
 parser.add_argument(
     "-F", "--error-format",
     default="{error.instance}: {error.message}\n",
@@ -43,6 +50,7 @@ def _json_file(path):
          "validators that are registered with jsonschema, simply the name "
          "of the class.",
 )
+
 parser.add_argument(
     "schema",
     help="the JSON Schema to validate with",
@@ -61,12 +69,18 @@ def main(args=sys.argv[1:]):
     sys.exit(run(arguments=parse_args(args=args)))
 
 
-def run(arguments, stdout=sys.stdout, stderr=sys.stderr):
+def run(arguments, stdin=sys.stdin, stdout=sys.stdout, stderr=sys.stderr):
     error_format = arguments["error_format"]
     validator = arguments["validator"](schema=arguments["schema"])
     errored = False
-    for instance in arguments["instances"] or ():
-        for error in validator.iter_errors(instance):
-            stderr.write(error_format.format(error=error))
+    instances = (json.loads(line) for line in stdin) if arguments.get("filter") else (arguments.get("instances") or ())
+    for instance in instances:
+        errors = tuple(validator.iter_errors(instance))
+        if errors:
             errored = True
+            for error in errors:
+                stderr.write(error_format.format(error=error))
+        elif arguments.get("filter"):
+            json.dump(instance, stdout, separators=(',', ':'))
+            stdout.write('\n')
     return errored

jsonschema/tests/test_cli.py

@@ -1,3 +1,5 @@
+import json
+
 from jsonschema import Draft4Validator, ValidationError, cli
 from jsonschema.compat import StringIO
 from jsonschema.tests.compat import mock, unittest
@@ -17,6 +19,11 @@ def iter_errors(self, instance):
     return FakeValidator
 
 
+def serialize_instances(*instances):
+    """:Return: a string with `instances` serialized to JSON, one per line."""
+    return '\n'.join(json.dumps(instance, separators=(',', ':')) for instance in instances) + '\n'
+
+
 class TestParser(unittest.TestCase):
     FakeValidator = fake_validator()
 
@@ -108,3 +115,31 @@ def test_unsuccessful_validation_multiple_instances(self):
         self.assertFalse(stdout.getvalue())
         self.assertEqual(stderr.getvalue(), "1 - 9\t1 - 8\t2 - 7\t")
         self.assertEqual(exit_code, 1)
+
+    def test_filter_valid(self):
+        serialized = serialize_instances(1, [2], {3: 4})
+        stdin, stdout, stderr = StringIO(serialized), StringIO(), StringIO()
+        exit_code = cli.run({
+            "validator": fake_validator(),
+            "schema": {},
+            "error_format": "{error.message}",
+            "filter": True,
+        }, stdin=stdin, stdout=stdout, stderr=stderr)
+        self.assertFalse(stderr.getvalue())
+        self.assertEquals(stdout.getvalue(), serialized)
+        self.assertEqual(exit_code, 0)
+
+    def test_filter_invalid(self):
+        schema = {"type": "object"}
+        serialized_in = serialize_instances({}, [2], {3: 4})
+        serialized_out = serialize_instances({}, {3: 4})
+        stdin, stdout, stderr = StringIO(serialized_in), StringIO(), StringIO()
+        exit_code = cli.run({
+            "validator": Draft4Validator,
+            "schema": schema,
+            "error_format": "{error.message}",
+            "filter": True,
+            }, stdin=stdin, stdout=stdout, stderr=stderr)
+        self.assertEquals(stderr.getvalue(), "[2] is not of type 'object'")
+        self.assertEquals(stdout.getvalue(), serialized_out)
+        self.assertEqual(exit_code, 1)