[Clarity] concat should allow to concatenate more than 2 sequences at once.

[LNow]

Is your feature request related to a problem? Please describe.
I want to concatenate more than 2 sequences at once without using fold because it is to expensive (execution costs wise).

Describe the solution you'd like
concat function should take 2 or more sequences of the same type and returns a concatenated sequence of the same type, with the resulting sequence_len = sequence1_len + sequence2_len ...+ ... sequenceN_len.

Instead of this:

(concat
  (concat
    (concat 0x00 0x01)
    (concat 0x02 0x03
  )
  (concat
    (concat 0x04 0x05)
    (concat 0x05 0x07)
  )
)
;; or
(fold my-concat (list 0x00 0x01 0x02 0x03 0x04 0x05 0x06 0x07) 0x)
(define-private (my-concat (input (buff 1)) (output (buff 8)))
  (unwrap-panic (as-max-len? (concat output input) u8))
)

I'd like to be able to do this:

(concat  0x00 0x01 0x02 0x03 0x04 0x05 0x06 0x07)

Describe alternatives you've considered
It is quite easy to concatenate multiple sequences using fold, but this approach is quite expensive.

[jcnelson]

Hey @LNow,

It is quite easy to concatenate multiple sequences using fold, but this approach is quite expensive.

What leads you to believe that (concat 0x00 0x01 0x02 0x03 0x04 0x05 0x06 0x07) will be significantly cheaper than the fold approach?

[LNow]

@jcnelson the answer to your question is quite simple. Current execution costs.

Using example I posted above.

Multiple nested concat:

+----------------------+----------+------------+
|                      | Consumed | Limit      |
+----------------------+----------+------------+
| Runtime              | 8337     | 5000000000 |
+----------------------+----------+------------+
| Read count           | 3        | 7750       |
+----------------------+----------+------------+
| Read length (bytes)  | 221      | 100000000  |
+----------------------+----------+------------+
| Write count          | 0        | 7750       |
+----------------------+----------+------------+
| Write length (bytes) | 0        | 15000000   |
+----------------------+----------+------------+

Fold:

+----------------------+----------+------------+
|                      | Consumed | Limit      |
+----------------------+----------+------------+
| Runtime              | 20130    | 5000000000 |
+----------------------+----------+------------+
| Read count           | 3        | 7750       |
+----------------------+----------+------------+
| Read length (bytes)  | 221      | 100000000  |
+----------------------+----------+------------+
| Write count          | 0        | 7750       |
+----------------------+----------+------------+
| Write length (bytes) | 0        | 15000000   |
+----------------------+----------+------------+

Of course the more elements we have to concat, multiple nested concat's consumes more Read length (bytes), but that is only because code becomes longer and longer. Runtime wise it is consistently 2x cheaper.

[jcnelson]

One thing we could do, regardless of how we address this, is make it so the Clarity VM does not re-run the type checks (and associated runtime costs) on each application of the function in fold.

[njordhov]

See also the proposal for variadic concat at clarity-lang/reference#17

[jcnelson]

PR #3294 is still in-flight and will be included in the next VM upgrade

[jcnelson]

Temporarily assigning this to @obycode. Please feel free to re-assign.