Add support for nulls in maps

[zsperske]

Putting this up to start a convo about a... bug (?) I'm noticing, not really sure. This fails to pass the delayed object check because when parsing as a JSONObject:

"myStringMapWithSingleNullValue": { "key1": null },

Once the parser reaches the {, the next call to JsonReader's peek method returns JsonToken.NULL. So we end up with a completely empty map. @ndtaylor is that expected? Is "key1" not a token?

[ndtaylor]

There are two problems that need to be addressed. One is that we're putting null into a JSONObject in parseAsJsonObject() and parseSpecificJsonObjectDelayed(), which basically removes the key in the JSONObject. Instead, we should put JSONObject.NULL as the value.

Second, in convertJsonObjectToMap(), jsonObject.opt(name) returns JSONObject.NULL if the value is null. We should use that as our signal to put null into the map we're generating.

[KennethNickles]

Can't maps have a single null key as well or is that a java specific implementation detail?

[ndtaylor]

Can't maps have a single null key as well or is that a java specific implementation detail?

I don't believe the JSON standard allows for a null key. Also, is this in response to the conversation in #6?

[KennethNickles]

Yeah sorry too many prs open.

[KennethNickles]

Zach told me to rescind this because of test failures.

[zsperske]

There are two problems that need to be addressed. One is that we're putting null into a JSONObject in parseAsJsonObject() and parseSpecificJsonObjectDelayed(), which basically removes the key in the JSONObject. Instead, we should put JSONObject.NULL as the value.

Second, in convertJsonObjectToMap(), jsonObject.opt(name) returns JSONObject.NULL if the value is null. We should use that as our signal to put null into the map we're generating.

Back from KotlinConf. Let me see I'm following. In parseAsJsonObject() we call parseNextValue(), are you talking about returning JSONObject.NULL instead of just null from case NULL:? Pushed that change up but it comes with a couple problems.

"myNullInt": null hits java.lang.NumberFormatException: For input string: "null" with this generated code:

out.myNullInt = Integer.valueOf(jsonObject.optString("myNullInt"));

And "key1": { "myString": null } parses the same as "key3": { "myString": "null" }

[ndtaylor]

In parseAsJsonObject() we call parseNextValue(), are you talking about returning JSONObject.NULL instead of just null from case NULL:? Pushed that change up but it comes with a couple problems.

This is what I meant.

[zsperske]

This is what I meant.

Ok cool, seems pretty similar to what I was doing. It still has the problems I listed above though. "Key1" parses to com.workday.autoparse.json.demo.SimpleTestObject@342c38f8[myString=post-parse:null,discrimValue=<null>]

[ndtaylor]

"Key1" parses to com.workday.autoparse.json.demo.SimpleTestObject@342c38f8[myString=post-parse:null,discrimValue=]

TestObject has an OnPostCreateChild() method that changes the value of myString. So it's working as intended.

[zsperske]

TestObject has an OnPostCreateChild() method that changes the value of myString. So it's working as intended.

I looked at that method, and it only does that if the String value isn't null. I would have expected it to be null and not have that bit added. I'll look into it more.

[ndtaylor]

I looked at that method, and it only does that if the String value isn't null.

Lol. Yup, there is something else going on there.

For the test that's failing because of a NumberFormatException, it's trying to convert null to an int, which doesn't work. I'd say it's questionable to allow null values in the JSON for primitives. If you (as a user of Autoparse) really want to allow for that, then I'd annotate a setter that takes a String, and have the setter handle the null value. Is this a use case you've actually run into?