[
https://issues.apache.org/jira/browse/THRIFT-5474?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17435733#comment-17435733
]
Grant Sherrick edited comment on THRIFT-5474 at 10/29/21, 1:21 AM:
-------------------------------------------------------------------
Ahhh, sorry about including {{TNotSoSimpleJSONProtocol}} in this description, I
needed to name my fork something in my local repo. That is my mistake. I'll
update the original description.
When I tested this against both jackson and PHP's json_decode, I get the same
error when trying to deserialize the String that is output from the above
thrift message and the resulting serialization using TSimpleJSONProtocol.
I tried to reproduce this by adding some jackson JSON parsing to each of the
tests for TSimpleJSONProtocol in my pull request. If the parsing fails, it will
throw a [JsonParseException|#createParser(java.lang.String)].
The issue here is that double quotes must be treated differently than other
escape characters because JSON values that are strings are surrounded in double
quotes, and these quotes are escaped a single time. The first double quote
surrounding the string itself is escaped a single time here and can be seen
[here|#L312].] However, the quotes that are internal to the string must be
escaped twice because otherwise they close the string itself. Does that make
sense to you? The string being produced here is being closed prematurely
because the internal quotes are not being escaped twice.
was (Author: thealmightygrant):
Ahhh, sorry about including {{TNotSoSimpleJSONProtocol}} in this description, I
needed to name my fork something in my local repo. That is my mistake. I'll
update the original description.
When I tested this against both jackson and PHP's json_decode, I get the same
error when trying to deserialize the String that is output from the above
thrift message and the resulting serialization using TSimpleJSONProtocol.
I tried to reproduce this by adding some jackson JSON parsing to each of the
tests for TSimpleJSONProtocol in my pull request. If the parsing fails, it will
throw a [JsonParseException|#createParser(java.lang.String)].]
The issue here is that double quotes must be treated differently than other
escape characters because JSON values that are strings are surrounded in double
quotes, and these quotes are escaped a single time. The first double quote
surrounding the string itself is escaped a single time here and can be seen
[here|#L312].] However, the quotes that are internal to the string must be
escaped twice because otherwise they close the string itself. Does that make
sense to you? The string being produced here is being closed prematurely
because the internal quotes are not being escaped twice.
> TSimpleJsonProtocol serialization on string with inner quotes does not
> produce valid JSON.
> ------------------------------------------------------------------------------------------
>
> Key: THRIFT-5474
> URL: https://issues.apache.org/jira/browse/THRIFT-5474
> Project: Thrift
> Issue Type: Bug
> Components: Java - Library
> Reporter: Grant Sherrick
> Priority: Minor
> Time Spent: 20m
> Remaining Estimate: 0h
>
> There is an issue in the generation of JSON from the TSimpleJSONProtocol,
> where if there are quotes inside of a string, then those quotes will not be
> serialized correctly. These quotes need one more layer of escaping to produce
> valid JSON.
>
> For instance, if you have the Thrift structure defined by:
>
> {code:java}
> struct MyThriftStruct {
> 1: required string my_string
> }
> {code}
>
> And, I then define an object of that type in Java:
> {code:java}
> MyThriftStruct t = new MyThriftStruct();
> t.my_string = "I said: \"this is a fairly nice string.\""{code}
>
> And, if I then serialize this using the TSimpleJsonProtocol in this manner:
> {code:java}
> TSerializer thriftJsonSer = new TSerializer(new
> TSimpleJSONProtocol.Factory());
> String myThriftStructJsonString = return thriftJsonSer.toString(t);
> {code}
>
> This will result in JSON that cannot be deserialized because the quotes
> inside of the string were not escaped properly.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
