Chained merges `<<:` don’t work

[danielbayley]

For example:

one: &a
  a: 1

two: &b
  <<: *a
  b: 2

three:
  <<: *b
  c: 3

should produce:

{
  "one": {
    "a": 1
  },
  "two": {
    "a": 1,
    "b": 2
  },
  "three": {
    "a": 1,
    "b": 2,
    "c": 3
  }
}

but instead, produces:

{
  "one": {
    "a": 1
  },
  "two": {
    "a": 1,
    "b": 2
  },
  "three": {
    "<<": {
      "a": 1
    },
    "b": 2,
    "c": 3
  }
}

[ingydotnet]

Good catch. Fixed in ffdf7f5.
Pushed to devel branch.
You can install ys from that branch with:

make install-ys PREFIX=</some/path/with/bin/in/PATH>  # defaults to $HOME/.local

Planning to release 0.1.86 tonight.
Looking at your PR #196 next.

[ingydotnet]

Fix ffdf7f5 released in 0.1.86

[danielbayley]

Nice one @ingydotnet, one thing remains though—so while this works now:

one: &a
  a: 1

two: &b
  <<: *a
  b: 2

three:
  <<: *b

(although the order of keys is reversed—which is probably a bug?)

{
  "one": {
    "a": 1
  },
  "two": {
    "b": 2,
    "a": 1
  },
  "three": {
    "c": 3,
    "b": 2,
    "a": 1
  }
}

This does not:

one: &a
  a: 1

two: &b
  <<: *a
  b: 2

three: *b

{
  "one": {
    "a": 1
  },
  "two": {
    "b": 2,
    "a": 1
  },
  "three": {
    "<<": {
      "a": 1
    },
    "b": 2
  }
}

[ingydotnet]

The key order is right afaik.
It's based on this old thing https://yaml.org/type/merge.html
Which says that you start with the mapping minus the << key, and then add keys from << if they don't exist.

Re the failing case, I suspected that when I fixed it but didn't push past your reported problem because I didn't have the time for it. I will take a look this week.

Please keep the issues coming.

[ingydotnet]

Looking at this now.

one: &a
  a: 1

two: &b
  <<: *a
  b: 2

three: *b

compiles to:

{"one" (_& 'a {"a" 1}),
 "two" (+merge (_& 'b {"<<" (_* 'a), "b" 2})),
 "three" (_* 'b)}

I need to have the _* function look for the merge key and do the merge.

There's a little problem though. At this point we don't know if the key was << '<<' or "<<".
The << has to be unquoted to be a YAML merge key.

I have an idea. I'm going to make the compiler produce

{"one" (_& 'a {"a" 1}),
 "two" (+merge (_& 'b {:-<< (_* 'a), "b" 2})),
 "three" (_* 'b)}

The :-<< is a valid Clojure keyword but something that will never show up as a YAML or YS data key.
I'm not even sure keywords make sense at all in data mode, but using a - in front just to be safe (in case it makes sense down the road).

[ingydotnet]

@danielbayley this is fixed in the devel branch now.
I'll likely release this on Sunday, but if you want to test things out you can install easily with: make install-ys.
It takes a minute or 2 to compile but doesn't require any deps.

[ingydotnet]

44f7fd5 is the fix commit, fwiw.

It was quite a refactoring of how this works but I think it makes it more solid now.

[ingydotnet]

Fixed released in 0.1.87

[danielbayley]

Fixed released in 0.1.87

Nice one @ingydotnet! Will check it out…

Although, with Homebrew being my preferred install method—actually the only method I allow as a strict self-imposed rule for everything on my machine—I'm wondering if there should be some CI step to automatically bump the formula version on release?

Edit: I notice there already is: the brew-update util (which, FYI, could also live as an external command over in the Homebrew tap in a cmd directory…), I think it just needs wiring up with an Actions workflow to trigger on: release…

[ingydotnet]

I barely know what I'm doing with homebrew. Any help you can offer would be great.

https://matrix.to/#/@ingy:yaml.io is a good way to chat with me directly if you want...

[danielbayley]

I barely know what I'm doing with homebrew. Any help you can offer would be great.

I can take a look at the setup… As it stands, the brew-update util kept failing for me, so I opened yaml/homebrew-yamlscript#1 as a manual update.

Note

See yaml/homebrew-yamlscript#3 @ingydotnet.

https://matrix.to/#/@ingy:yaml.io is a good way to chat with me directly if you want...

Cool, I'm really hesitant to sign up for yet another account for anything though — according to 1Password I already have well over 400! Trying to cut down…