-
Type:
Bug
-
Resolution: Done
-
Priority:
Major - P3
-
Affects Version/s: 3.0.0, 3.1.0
-
Component/s: bsondump, mongoexport
-
None
-
Environment:Linux x86_64
-
Kernel Tools 8 08/28/15
-
v3.0
bsondump and mongoexport can't handle BSON Regexps (type 0x0B) with patterns that json.RegExp doesn't understand. Since the BSON spec states that Regexp patterns can be any cstring (null-terminated string of non-null bytes), these are valid BSON (and might be valid regexes for other regex engines).
$ cat tools-regex-fail.json
{ "_id" : { "$oid" : "51193479ef23779f3d000cbd" }, "keywords_regex" : { "$regex" : "\\.", "$options" : "" } }
$ /m/3.0.4/bin/mongoimport --port 6834 -d test -c test3 --drop tools-regex-fail.json
2015-07-25T22:04:26.488+1000 connected to: localhost:6834
2015-07-25T22:04:26.488+1000 dropping: test.test3
2015-07-25T22:04:26.490+1000 imported 1 document
$ /m/3.0.4/bin/mongodump --port 6834 -d test -c test3 -o tools-regex-fail-dump
2015-07-25T22:06:36.283+1000 writing test.test3 to tools-regex-fail-dump/test/test3.bson
2015-07-25T22:06:36.284+1000 writing test.test3 metadata to tools-regex-fail-dump/test/test3.metadata.json
2015-07-25T22:06:36.285+1000 done dumping test.test3
$ /m/3.0.4/bin/bsondump tools-regex-fail-dump/test/test3.bson
2015-07-25T22:09:49.651+1000 unable to dump document 1: error converting doc to JSON: json: error calling MarshalJSON for type json.RegExp: invalid character '.' in string escape code
2015-07-25T22:09:49.652+1000 1 objects found
$ /m/3.0.4/bin/mongoexport --port 6834 -d test -c test3
2015-07-25T22:12:04.947+1000 connected to: localhost:6834
2015-07-25T22:12:04.948+1000 Failed: json: error calling MarshalJSON for type json.RegExp: invalid character '.' in string escape code
$ /m/3.1.6/bin/bsondump tools-regex-fail-dump/test/test3.bson
2015-07-25T22:41:18.989+1000 unable to dump document 1: error converting doc to JSON: json: error calling MarshalJSON for type json.RegExp: invalid character '.' in string escape code
2015-07-25T22:41:18.989+1000 1 objects found
$ /m/3.1.6/bin/mongoexport --port 6834 -d test -c test3
2015-07-25T22:41:27.804+1000 connected to: localhost:6834
2015-07-25T22:41:27.805+1000 Failed: json: error calling MarshalJSON for type json.RegExp: invalid character '.' in string escape code
This is a regression from the 2.6 tools, which handle such regexes without problem. Which means that this is the workaround (unless a 3.0 feature is required, eg. SCRAM-SHA-1).
$ /m/2.6.10/bin/bsondump tools-regex-fail-dump/test/test3.bson
{ "_id" : ObjectId( "51193479ef23779f3d000cbd" ), "keywords_regex" : /\\./ }
1 objects found
$ /m/2.6.10/bin/mongoexport --port 6834 -d test -c test3
connected to: 127.0.0.1:6834
{ "_id" : { "$oid" : "51193479ef23779f3d000cbd" }, "keywords_regex" : { "$regex" : "\\.", "$options" : "" } }
exported 1 records
Other BSON types may have similar problems, if native golang objects can only represent a subset of the valid BSON values. (Looking over the other types, this seems unlikely.)
There may also be similar problems in the other tools (although I haven't found any).