Implement batch_all and update Utility pallet for weight refunds

[xlc]

Fixes #6519

Also updated #[transactional] to make it handle returns correctly.

polkadot companion: paritytech/polkadot#1775

[bkchr]

Looks good, but not sure if we need some different base weight for transactional dispatchables?

@shawntabrizi

[shawntabrizi]

I don't see why there would be different weights based on transactional dispatch. Should there be?

@xlc can you confirm quickly that the base weight running batch and batch_all are the same on your local computer?

If you want to be really safe, it is probably best for each function to get its own weight function anyway, rather than sharing one. It allows more flexibility in the future for customization and for the functions to potentially diverge in logic.

[xlc]

Local benchmark result (with bunch browser tabs open and IDE running)

Pallet: "pallet_utility", Extrinsic: "batch", Lowest values: [], Highest values: [], Steps: [], Repeat: 200
Median Slopes Analysis
========
-- Extrinsic Time --

Model:
Time ~=    239.7
    + c    0.461
              µs

Reads = 0 + (0 * c)
Writes = 0 + (0 * c)
Min Squares Analysis
========
-- Extrinsic Time --

Data points distribution:
    c   mean µs  sigma µs       %
    0     237.6     2.219    0.9%
  100     313.2     20.86    6.6%
  200     340.2     11.81    3.4%
  300     378.1     3.134    0.8%
  400       423     3.532    0.8%
  500     477.3     16.24    3.4%
  600       591     6.071    1.0%
  700     612.9     38.86    6.3%
  800     627.1     28.18    4.4%
  900     655.1     4.802    0.7%
 1000     697.9     0.868    0.1%

Quality and confidence:
param     error
c         0.002

Model:
Time ~=    251.8
    + c     0.47
              µs

Reads = 0 + (0 * c)
Writes = 0 + (0 * c)
Pallet: "pallet_utility", Extrinsic: "batch_all", Lowest values: [], Highest values: [], Steps: [], Repeat: 200
Median Slopes Analysis
========
-- Extrinsic Time --

Model:
Time ~=    255.1
    + c    0.448
              µs

Reads = 0 + (0 * c)
Writes = 0 + (0 * c)
Min Squares Analysis
========
-- Extrinsic Time --

Data points distribution:
    c   mean µs  sigma µs       %
    0     253.5     2.751    1.0%
  100     299.2      0.42    0.1%
  200     348.8     3.611    1.0%
  300     390.2     4.021    1.0%
  400     471.6     15.69    3.3%
  500     479.1     23.33    4.8%
  600     521.5     6.381    1.2%
  700     568.6     16.76    2.9%
  800       599     0.832    0.1%
  900     669.9     22.94    3.4%
 1000     744.6     18.76    2.5%

Quality and confidence:
param     error
c         0.001

Model:
Time ~=    254.3
    + c    0.463
              µs

diff --git a/frame/utility/src/benchmarking.rs b/frame/utility/src/benchmarking.rs
index 8ca0e216f..8713fecae 100644
--- a/frame/utility/src/benchmarking.rs
+++ b/frame/utility/src/benchmarking.rs
@@ -49,6 +49,19 @@ benchmarks! {
                assert_last_event::<T>(Event::BatchCompleted.into())
        }

+       batch_all {
+               let c in 0 .. 1000;
+               let mut calls: Vec<<T as Trait>::Call> = Vec::new();
+               for i in 0 .. c {
+                       let call = frame_system::Call::remark(vec![]).into();
+                       calls.push(call);
+               }
+               let caller = whitelisted_caller();
+       }: _(RawOrigin::Signed(caller), calls)
+       verify {
+               assert_last_event::<T>(Event::BatchCompleted.into())
+       }
+
        as_derivative {
                let u in 0 .. 1000;
                let caller = account("caller", u, SEED);

[shawntabrizi]

@xlc and @bkchr please take a look at and review my changes.

Note that your PR actually had a mistake, where you assume weight should start at zero, but it should actually start at base_weight using the WeightInfo.

[bkchr]

LGTM

Does this needs an audit?

[shawntabrizi]

Does this needs an audit?

I have added the nice to have audit tag, but it seems the only logic it is touching is weights, with no so much concern from me?

@gavofyork what do you think?

[xlc]

ping @gavofyork

[viniul]

We are done with the review for this PR - looks good to us!

[bkchr]

@xlc please merge master and we can merge this! :)

[shawntabrizi]

bot merge

[ghost]

Trying merge.