Fix path wildcards in combination with arrays.

[blackwinter]

Resolves #115 and #119.

NOTE: The updated test MetafixRecordTest.appendWithAsteriksWildcardAtTheEnd() was probably wrong in the first place; the original comment expected the other values as well, but the actual review (which was the basis for the unit test) did not.

[fsteeg]

All these enabled test cases 🙌

[TobiasNx]

Resolves #115 and #119.

NOTE: The updated test MetafixRecordTest.appendWithAsteriksWildcardAtTheEnd() was probably wrong in the first place; the original comment expected the other values as well, but the actual review (which was the basis for the unit test) did not.

I have checked that test, it is not wrong the new output seems wrong imo.

The comment suggested/asked how to handle asterisk at the end of an wild-card if it does include all variations of the whole path incl. the subfields.

Now the behaviour seems weird.

input is this:

{
	"animals": ["dog", "cat", "zebra"],
	"animal": "bunny",
	"animols": {
		"name": "bird",
		"type": "TEST"
	},
	"ANIMALS": "dragon and unicorn"

}

Fix:

copy_field("ani*", "stringimals[].$append")

set_array("strongimals[]")
copy_field("ani*", "strongimals[].$append")

ani* => as expecded would only match with the field "animal"

my comment asked if the behaviour of fix would be the same as morph. then it would match with "animals", "animal", "animols" and include other values.

Now the result includes "animal" and "animals" and transforms them to simple string values and does some strange behaviour from the other tests since it does not provide index numbers in "stringimals" and adds $append.

If "animals", "animal" and "animols" should be included, so not just simple string-fields but also objects and arrays, the expected would be:

  "stringimals" : {
    "1" : ["dog", "cat", "zebra"],
    "2" :  "bunny",
    "3" : {"name": "bird", "type": "TEST"}
  } ,
  "strongimals" : [ ["dog", "cat", "zebra"], "bunny" ,  {"name": "bird", "type": "TEST"}]

[blackwinter]

my comment asked if the behaviour of fix would be the same as morph. then it would match with "animals", "animal", "animols" and include other values.

No, those are two different things:

1. Do Fix path wildcards match the behaviour of Morph wildcards in that they extend to the whole path?
2. What happens if a Fix path wildcard matches an array or a hash?

As for 1., we (well, I) decided against it (see my answer to your comment); Fix path wildcards only ever match a single path segment.

The behaviour for 2. is what changed here. Previously, the array (animals[]) would simply get stuck in a dead end, that's why IMO the test was wrong.

The pattern ani* matches all field names animals[], animal and animols, so it's basically equivalent to

copy_field("animals[]", "stringimals[].$append")
copy_field("animal", "stringimals[].$append")
copy_field("animols", "stringimals[].$append")

The question then becomes whether each of these cases is implemented correctly (or at all). And copy_field("array", "new_array[].$append") already works, that's not new (see unit test on master: 46ced0b). Only the last case (copy_field("hash", "new_array[].$append)) does not work at the moment.

[TobiasNx]

my comment asked if the behaviour of fix would be the same as morph. then it would match with "animals", "animal", "animols" and include other values.

No, those are two different things:

1. Do Fix path wildcards match the behaviour of Morph wildcards in that they extend to the _whole_ path?

2. What happens if a Fix path wildcard matches an array or a hash?

1. Yes that was the reason I modeled the test like this and not include subfields incl. array-fields.
2. Agreed. Also agree that the object/hash handling is missing.

But also how array and hash are handled is relevant.

But this could be a problem of the fix and not of your changes.
(It is a general problem: https://metafacture.org/playground/?flux=PG_DATA%0A%7Cas-records%0A%7Cdecode-json%0A%7Cfix%0A%7Cencode-json%28prettyPrinting%3D%22true%22%29%0A%7Cprint%0A%3B&fix=copy_field%28%22a%5B%5D%22%2C+%22z%5B%5D.%24append%22%29&data=%7B%0A++++%22a%22%3A+%5B%22b%22%2C+%22c%22%2C+%22d%22%5D%2C%0A++++%22z%22%3A+%5B%5D%0A%7D&active-editor=fix)

copy_field("animals[]", "stringimals[].$append")

operates like I would expect copy_field("animals[].*", "stringimals[].$append")

[blackwinter]

But this could be a problem of the fix and not of your changes.

Exactly, this is independent of the changes discussed here as it's already working this way.

[TobiasNx]

But this could be a problem of the fix and not of your changes.

Exactly, this is independent of the changes discussed here as it's already working this way.

my questions are solved. +1
I opened another ticket for the array handling metafacture/metafacture-core#604