Schema parsing / conform

[nilern]

Something like this for #241 and the rest of #180?

Builds on top of #317.

• On failure, (m/conform s v) returns :malli.core/nonconforming
• For most schemas (m/conform s v) just returns something that is = to v (optimally, v itself).
• But for the (new) :or* (tagged :or) (= (m/conform [:or* [:s string?] [:n int?]] 0) (miu/-tagged :n 0))
  • Non-identity conforming like this propagates: (= (m/conform [:tuple [:or* [:s string?] [:n int?]]] [0]) [(miu/-tagged :n 0)])
• Seqexes conform much as in Spec:

(m/conform [:alt string? int?] [0]) ; => 0
(m/conform [:alt* [:s string?] [:n int?]] [0]) ; => (miu/-tagged :n 0)

(m/conform [:cat [:* string?] [:* int?]] ["foo" "bar" 1 2]) ; => [["foo" "bar"] [1 2]]
(m/conform [:cat* [:s [:* string?]] [:n [:* int?]]] ["foo" "bar" 1 2]) ; => {:s ["foo" "bar"], :n [1 2]}

(m/conform [:? string?] ["foo"]) ; => "foo"
(m/conform [:? string?] []) ; => nil

(m/conform [:* string?] ["foo" "bar"]) ; => ["foo" "bar"]
(m/conform [:* string?] []) ; => []

(m/conform [:+ string?] ["foo" "bar"]) ; => ("foo" "bar")
(m/conform [:+ string?] []) ; => :malli.core/nonconforming

(m/conform [:repeat {:min 1, :max 3} string?] []) ; => :malli.core/nonconforming
(m/conform [:repeat {:min 1, :max 3} string?] ["foo" "bar"]) ; => ["foo" "bar"]
(m/conform [:repeat {:min 1, :max 3} string?] ["foo" "bar" "baz" "quux"]) ; => :malli.core/nonconforming

(m/conform [:cat* [:s [:schema [:* string?]]]] [["foo" "bar"]]) ; => {:s ["foo" "bar"]}

[nilern]

At this point it should be quite simple to adjust the result format to taste.

[nilern]

Also, match, parse or even destructure would be a more usual name, especially if it is not going to be as similar to Spec as it is now.

[ikitommi]

This is really good. Few suggestions:

1. the conform should not throw, should fail with ::m/conforming instead, have a code for this (on top of this PR), makes the backtracking order of magnitude faster
2. still think that we should use explain for conform, the code looks like 90% of c&p now. Can be done later
3. the conform, I suggest parse, can do a PR on top of this.