fixes the macro expansion of $> and $< specs

[joemfb]

The expansions of $> and $< were infinitely recursive. These are new specs for building molds from parts of other $% molds. They are ?>-style assertions against the head of the noun being cast or nest-checked. For example:

> `task:able:ames`[%west ~zod /foo 1]
[%west p=~zod q=/foo r=1]
>
> `$>(%west task:able:ames)`[%west ~zod /foo 1]
[%west p=~zod q=/foo r=1]
> 
> `$<(%west task:able:ames)`[%west ~zod /foo 1]
nest-fail

I think it would be very helpful to use these specs when building the +note and +sign molds for vanes, which are obviously subtypes of the +task and +gift types of the other vanes in the system. If there's interest, I'd be happy to do so now.

[joemfb]

These specs have some limitations that should be called out explicitly. Most notably, some of them are not directly bunt-able:

> *$<(%west task:able:ames)
[%want p=~zod q=/ r=0]
> *$>(%west task:able:ames)
ford: %ride failed to execute:

This "intuitively" follows from that fact that specs are not available at runtime, but the user-experience is less than intuitive. The ability-to-bunt can be restored (in a frustratingly-explicit nanner) with $~:

> *$~([%west ~zod / 0] $>(%west task:able:ames))
[%west p=~zod q=/ r=0]

As I mentioned, I think the right way to use these specs is in building up a $% that references other specs, like +note and +sign in the vanes. The whole $% should be wrapped in a single $~ to ensure that it bunts, but the individual "pages" of the tagged-union "book" need not be bunt-able.

[belisarius222]

LGTM

[cgyarvin]

Ha not surprised that I f*ed this up. Yes these are very useful for dynamic checks of static properties we can’t statically check. They should be applied to containers as well — more urgent. Also to a layer of managed string types. No one should be able to send you a misordered set or a dodgy string over the network.

…

Sent from my iPhone

On May 24, 2019, at 2:36 PM, Joe Bryan ***@***.***> wrote: The expansions of $> and $< were infinitely recursive. These are new specs for building molds from parts of other $% molds. They are ?>-style assertions against the head of the noun being cast or nest-checked. For example: > `task:able:ames`[%west ~zod /foo 1] [%west p=~zod q=/foo r=1] > > `$>(%west task:able:ames)`[%west ~zod /foo 1] [%west p=~zod q=/foo r=1] > > `$<(%west task:able:ames)`[%west ~zod /foo 1] nest-fail I think it would be very helpful to use these specs when building the +note and +sign molds for vanes, which are obviously subtypes of the +task and +gift types of the other vanes in the system. If there's interest, I'd be happy to do so now. You can view, comment on, or merge this pull request online at: #1176 Commit Summary fixes the macro expansion of $> and $< specs File Changes M sys/hoon.hoon (4) Patch Links: https://github.com/urbit/arvo/pull/1176.patch https://github.com/urbit/arvo/pull/1176.diff — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[joemfb]

I think $| is the key for dynamic checks. +map should be something like:

+*  map  [key value]
  $|  (tree (pair @ @))
  |=(a=(tree (pair @ @)) ~(apt by a))

For example:

> ($|((tree (pair @ @)) |=((tree (pair @ @)) ~(apt by +<))) [[1 2] ~ [[3 4] ~ ~]])
{[p=1 q=2] [p=3 q=4]}
> ($|((tree (pair @ @)) |=((tree (pair @ @)) ~(apt by +<))) [[1 2] [[3 4] ~ ~] ~])
ford: %ride failed to execute:

[cgyarvin]

There is also a way to fix this...

…

Sent from my iPhone

On May 24, 2019, at 3:28 PM, Joe Bryan ***@***.***> wrote: These specs have some limitations that should be called out explicitly. Most notably, some of them are not directly bunt-able: > *$<(%west task:able:ames) [%want p=~zod q=/ r=0] > *$>(%west task:able:ames) ford: %ride failed to execute: This "intuitively" follows from that fact that specs are not available at runtime, but the user-experience is less than intuitive. The ability-to-bunt can be restored (in a frustratingly-explicit nanner) with $~: > *$~([%west ~zod / 0] $>(%west task:able:ames)) [%west p=~zod q=/ r=0] As I mentioned, I think the right way to use these specs is in building up a $% that references other specs, like +note and +sign in the vanes. The whole $% should be wrapped in a single $~ to ensure that it bunts, but the individual "pages" of the tagged-union "book" need not be bunt-able. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[cgyarvin]

Ha ha of course, wrong rune pair.

…

Sent from my iPhone

On May 24, 2019, at 4:09 PM, Joe Bryan ***@***.***> wrote: I think $| is the key for dynamic checks. +map should be something like: +* map [key value] $| (tree (pair @ @)) |=(a=(tree (pair @ @)) ~(apt by a)) For example: ($|((tree (pair @ @)) |=((tree (pair @ @)) ~(apt by +<))) [[1 2] ~ [[3 4] ~ ~]]) {[p=1 q=2] [p=3 q=4]} ($|((tree (pair @ @)) |=((tree (pair @ @)) ~(apt by +<))) [[1 2] [[3 4] ~ ~] ~]) ford: %ride failed to execute: — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

[philipcmonk]

Ooh, very nice. I would like to see the vanes moved to this style soon.