validate map/set treap order with $|

[joemfb]

after:

> ((map) [[1 2] ~ [[3 4] ~ ~]])
{[p=1 q=2] [p=3 q=4]}
> ((map) [[1 2] [[3 4] ~ ~] ~])
ford: %ride failed to execute:

before:

> ((map) [[1 2] ~ [[3 4] ~ ~]])
{[p=1 q=2] [p=3 q=4]}
> ((map) [[1 2] [[3 4] ~ ~] ~])
{[p=3 q=4] [p=1 q=2]}

much before:

#311

I had to move these mold/mold-builder definitions so that the +in and +by cores are available to them. Note that the section labels are still wrong, although somewhat less so. I think that's out of scope for this PR, as the labels are wrong in many parts of hoon.hoon.

[jtobin]

LGTM!

[philipcmonk]

Nice! When does the validator get run? Only on clam, or also on cast? Can you take a well-formed map, perturb it maliciously, and cast the result to map?

Approving because regardless of the above question, this is better than what we had.

[belisarius222]

^- and ^+ casting do not clam, so the menace of malicious map misordering remains. At least now the validator will get called on network data though, so you have to be on the inside (e.g. an app) to cause a problem. — ~rovnys-ricfer https://urbit.org

…

On Wed, May 29 2019 at 10:32 AM, < ***@***.*** > wrote: ***@***.**** approved this pull request. Nice! When does the validator get run? Only on clam, or also on cast? Can you take a well-formed map, perturb it maliciously, and cast the result to map? Approving because regardless of the above question, this is better than what we had. — You are receiving this because your review was requested. Reply to this email directly, view it on GitHub ( #1180?email_source=notifications&email_token=AAGVR5NWSX3MMSUBB7E2SJ3PX24Z7A5CNFSM4HQJSQG2YY3PNVWWK3TUL52HS4DFWFIHK3DMKJSXC5LFON2FEZLWNFSXPKTDN5WW2ZLOORPWSZGOB2BCROI#pullrequestreview-243411129 ) , or mute the thread ( https://github.com/notifications/unsubscribe-auth/AAGVR5IX4GURMV35VMVLO33PX24Z7ANCNFSM4HQJSQGQ ).

[joemfb]

Right, molds (specs) only exist at compile time. ^-(spec hoon) desugars to ^+(*spec hoon), so only the bunt of the mold is used for casting -- the validation function is not involved (or even available).

Some more examples:

> ((set) [2 [1 ~ ~] ~])
{1 2}
> ((set) [2 ~ [1 ~ ~]])
ford: %ride failed to execute:

Note also that the validation is applied recursively ((jug) is (map * (set))):

> `*`(my [1 (sy 2 3 ~)] [4 (sy 5 6 ~)] ~)
[[4 6 [5 0 0] 0] [[1 2 0 3 0 0] 0 0] 0]
> ((jug) [[4 [6 [5 ~ ~] ~]] [[1 [2 ~ [3 ~ ~]]] ~ ~] ~])
{[p=1 q={2 3}] [p=4 q={5 6}]}
> ((jug) [[4 [6 [5 ~ ~] ~]] ~ [[1 [2 ~ [3 ~ ~]]] ~ ~]])
ford: %ride failed to execute:
> ((jug) [[4 [6 ~ [5 ~ ~]]] [[1 [2 ~ [3 ~ ~]]] ~ ~] ~])
ford: %ride failed to execute:

[philipcmonk]

Makes sense.

Is there a reason why we don't verify the vertical ordering of a qeu?

[joemfb]

I guess nobody has taken the time to write +apt:to. I think it would be something like this:

++  apt
  ^-  ?
  ?~  a  &
  |-  ^-  ?
  ?~  l.a  &
  ?~  r.a  &
  &((mor n.l.a n.r.a) $(a l.a) $(a r.a))

[jtobin]

(extra 👍 for the tests)