simplify paths/0 and paths/1

[asheiduk]

recurse/0 already handles traversing objects and arrays, so it is more consistent to use that.
For paths/1 it is easier to use the actual value returned by recurse instead of querying that value with getpath/1 later.

[emanuele6]

Could even use .. instead of recurse.
.. is syntax sugar that compiles to a call to recurse/0.

[nicowilliams]

LGTM.

[nicowilliams]

@emanuele6 did you want to require the use of ..?

[wader]

LGTM but i noticed there is no tests for paths/1, time to add?

[asheiduk]

@wader There is something in man.test:

[paths(type == "number")]
[1,[[],{"a":2}]]
[[0],[1,1,"a"]]

[emanuele6]

@emanuele6 did you want to require the use of ..?

Well, I thought it would look neater with .....

Also, it would be easier to read since path(..) is a common pattern, and .. is commonly used, while recurse with no arguments is basically never used, but fine either way I guess.

[asheiduk]

@emanuele6 did you want to require the use of ..?

Well, I thought it would look neater with .....

Also, it would be easier to read since path(..) is a common pattern, and .. is commonly used, while recurse with no arguments is basically never used, but fine either way I guess.

@emanuele6 I think this is true for "user" code but here the definition of recurse is in the same file -- a file internal to jq itself. Having this textual direct link (e.g. via double-clicking) seemed more important to me. But that's only my take.

After all: Shall I change it so the PR can be merged or was this only an optional suggestion?

[emanuele6]

Well, it is fine with recurse

[emanuele6]

Thank you!