Make JSONPath results more consistent

[sgreben]

Background

Currently, our version of JSONPath has two index-selection behaviors:

Union-selection

• input: [[1,2],["a","b"]]
• path: [*][0,1]
• output: [1,"a",2,"b"]]

The results are concatenated: first all 0-components, then all 1-components. This is the default behavior for field selection .[myField1, myField2] and was also chosen for index-selection for consistency.

All-selection

• input: [[1,2],["a","b"]]
• path: [*][*]
• output: [1,2,"a","b"]

The results are interleaved: 0,1,0,1. This is the default JSONPath index selection behavior.

Proposal

Having both behaviors is confusing, all of these should have the same result.

Furthermore, it seems like echo '[[0,123],[1,456]]' | jp -xy '[*]', or echo '[[0,123],[1,456]]' | jp -xy '$', or even echo '[[0,123],[1,456]]' | jp should produce a valid plot:

 479.52│                     ▗
       │                  ▗▄▀▘
       │               ▄▄▀▘   
       │            ▄▞▀       
       │         ▄▞▀          
       │      ▄▞▀             
       │  ▗▄▀▀                
       │▄▀▘                   
 116.55└──────────────────────
       0                     1

Currently, only jp -xy '[*][0,1]' produces this.

I would propose that

1. Results are always interleaved (i.e. [*][*] = [*][0,1] = 0.0, 0.1, 1.0, 1.1)
2. An array of arrays is a valid result of the -xy JSONpath.
3. The default behavior (no JSONPaths specified) is to assume [[x,y],[x,y],...] input.

Proposal 3. in particular is quite debatable, so I'd appreciate any feedback on this.

[sgreben]

This has been addressed in 1.1.9, but I'll keep the issue open in case someone has an opinion :)

[sgreben]

dead issue cleanup