Poor Error consistency for conversion to LabVIEW types

[James Powell]

JSONtext has poor consistency of what happens when there is a mismatch between the JSON and the LabVIEW type.

Examples (see “Null treatment.vi”):

• null → integer: error 4012842
• true → integer: error 4012841
• null → string: error 4012841
• any scalar → cluster: no error, default
• any scalar → array: error 402840, empty array (even if default not empty)

Related issues:

• should null-->”” if a string?
• should a scalar be equivalent to a one-element array?

Also, some users want the option of treating null as default, without error (should this option override null-->NaN?)

See also issues #7, #47, #51

[James Powell]

Also enums:

• null/true/123 → enum error 4012841
• “<not in enum>“ → enum: error 4012843

‌

‌

[James Powell]

Current codes are:

<nierror code="402840">
JSONtext: Input JSON has wrong type for requested operation

<nierror code="402841">
JSONtext: Parsing Error. Invalid JSON formating (example: missing coma or quote) or unexpected type (example: found an Array when an Object was required)

<nierror code="402842">
JSONtext: Error

<nierror code="402843">
JSONtext: LabVIEW Type not implemented with JSONtext library.

‌

So really we should be using 402840 mostly.

[James Powell]

Actually, 402840 is used for things like trying to insert an a non-object, so perhaps a new error code:

402844: unable to convert JSON Value to supplied LabVIEW Type

[James Powell]

Added new error, and better message for all scaler types.

[James Powell]

Fixed failure with Fixed Point and null (made gve error).

Fixed issue where null convert to Cluster gives no error.

[James Powell]

Added to Test Cases

[James Powell]

Note: had to extensively rework the to Cluster bit of teh code (needed it). Now clearer how it handles Cluster to Array (Ragged Array, or unnamed cluster).

[James Powell]

bd30313503daad01c157ffc4052c2c129e82401c