Invalid JSON breaks custom query builder

[michael@…]

I just upgraded from 0.11.6 to 0.12.4 (and now 0.12.5dev-r11368) and am now getting a javascript syntax error on the custom query page. The offending javascript is the following:

var modes={"id":[{"name":"is","value":},{"name":"is not","value":"!"}],"select":[{"name":"is","value":},{"name":"is not","value":"!"}],"text":[{"name":"contains","value":"~"},{"name":"doesn't contain","value":"!~"},{"name":"begins with","value":"^"},{"name":"ends with","value":"$"},{"name":"is","value":},{"name":"is not","value":"!"}],"textarea":[{"name":"contains","value":"~"},{"name":"doesn't contain","value":"!~"}]};

To be sure, I check the generated JSON on the Demo site running Trac 0.12.5dev-r11311. Sure enough, that JSON is valid:

var modes={"id":[{"name":"is","value":""},{"name":"is not","value":"!"}],"select":[{"name":"is","value":""},{"name":"is not","value":"!"}],"text":[{"name":"contains","value":"~"},{"name":"doesn't contain","value":"!~"},{"name":"begins with","value":"^"},{"name":"ends with","value":"$"},{"name":"is","value":""},{"name":"is not","value":"!"}],"textarea":[{"name":"contains","value":"~"},{"name":"doesn't contain","value":"!~"}]};

I wonder if this bug is related to the version of python (2.4.3) I am running, and is similar in scope to #10771.

json.txt

@@ -1 @@
-var modes={"id":[{"name":"is","value":""},{"name":"is not","value":"!"}],"select":[{"name":"is","value":""},{"name":"is not","value":"!"}],"text":[{"name":"contains","value":"~"},{"name":"doesn't contain","value":"!~"},{"name":"begins with","value":"^"},{"name":"ends with","value":"$"},{"name":"is","value":""},{"name":"is not","value":"!"}],"textarea":[{"name":"contains","value":"~"},{"name":"doesn't contain","value":"!~"}]};
+var modes={"id":[{"name":"is","value":},{"name":"is not","value":"!"}],"select":[{"name":"is","value":},{"name":"is not","value":"!"}],"text":[{"name":"contains","value":"~"},{"name":"doesn't contain","value":"!~"},{"name":"begins with","value":"^"},{"name":"ends with","value":"$"},{"name":"is","value":},{"name":"is not","value":"!"}],"textarea":[{"name":"contains","value":"~"},{"name":"doesn't contain","value":"!~"}]};

[Christian Boos]

Damn, you're right, we've made a blunder here.

trac/util/text.py

@@ -114 +114 @@
     (conform to the JSON spec)
     """
     if not text:
-        return ''
+        return '""'
     def replace(match):
         return _js_quote[match.group(0)]
     return '"%s"' % _js_string_re.sub(replace, text)

trac/util/tests/text.py

@@ -77 +77 @@
                          to_js_string('\x02\x1e'))
         self.assertEqual(r'"\u0026\u003c\u003e"',
                          to_js_string('&<>'))
+        self.assertEqual('""',
+                         to_js_string(''))
+        self.assertEqual('""',
+                         to_js_string(None))
 
 
 class UnicodeQuoteTestCase(unittest.TestCase):

[anonymous]

Yup, that does it!

[Christian Boos]

Applied in r11382.

[Jun Omae]

Oooh. Sorry, I'm so stupid…. Thanks the reporter and Christian!!

[Christian Boos]

Replying to jomae:

Oooh. Sorry, I'm so stupid…. Thanks the reporter and Christian!!

Well, I get most of the blame, given all the iterations I triggered… and my final review saying "perfect" ;-)

[dale.miller@…]

I opened #10982 because this also needs to be done for 1.1, 1.0.x, and 1.1.x.