Rationale: in practice, a ProofId was always a slot id, and
an EndId was always a request id. Now that the definitons
of SlotId and RequestId are separated from the Marketplace,
we can import and use them.
- Marketplace tests for requestsForHost, and additional tests for myRequests and mySlots
- Added Utils lib with tests
- Added additional Bytes32AddressSetMap.keys expectations
Add `Bytes32AddressSetMap` which maps addresses to a requestId. This is used in `Marketplace.activeRequestsForHost`, where all addresses for a particular requestId are listed. This can then be used to iterate and list out the actives requests for a particular address in a view function only. This allows for all addresses for a request to be cleared in situations such as when a request fails or is cancelled.
Reinstate the index mapping for `activeSlots`, however with a modified data structure. The main difference comes from a change in `mySlots`: it now requires a `RequestId` as a parameter (which can be obtained using `myRequests`. This allowed for the `activeSlots` mapping to be keyed on `RequestId`, which gives a mapping of `ActiveSlotId => EnumerableSet.Bytes32`. `ActiveSlotId` is a keccak hash of `address + activeSlotIdx`, which not only allows retrieval of all active slot ids per request, but also allows clearing of all active slot ids for a request by incrementing the `activeSlotIdx`.
```
requestId => [ keccak(address + activeSlotIdx) => EnumerableSet.Bytes32 ]
```
Instead, iterate all active slots and remove them individually, only if they are part of a specific request. This is only for the case of request failure.
Allow for clearing of active slots by host, by incrementing an mapping index. The new index points to a fresh instance of EnumerableSet, effectively wiping it clean.
There may be cases where the the request end is not accurate as the state of the request hasn’t yet been updated. For example, when a request is cancelled, the request end would not have been updated to be in the past, and would still be set for the end of the request (which could be in the future).
1. Remove proof extension as it is not needed. Host are required to provide proofs from the moment they fill a slot, for the duration specified by the contract. This means that the ending of their requirements will be staggered at the end, as they were at the start, but this is more predicable for determining the cost of a request.
2. The proof end time was modified so that if the request state is not accepting proofs, it takes the min of the slot proof end time, the request end time, or block.timestamp - 1, which ensures that it returns a time in the past. If the slot is accepting proofs, it returns the slot end time.
3. Modify marketplace tests so that `waitUntilFinished` advances time to the proof ending of the last slot filled.
1. Replace all instances of `now()` with `await currentTime()` to get a more accurate representation of time from the block timestamp. Update examples.js to be async.
2. Move `RequestState` to `marketplace.js`
3. Delete `TestStorage` as `slashAmount` function no longer needed.
Allow proof ending to be extending once a contract is started, so that all filled slots share an ending time that is equal to the contract end time. Added tests.
Add a mapping for proof id to endId so that proof expectations can be extended for all proofs that share a given endId.
Add function modifiers that require the request state to allow proofs, with accompanying tests.
General clean up of each function’s request state context, with accompanying tests.
General clean up of all tests, including state change “wait” functions and normalising the time advancement functions.
Add function `isFinished` that checks whethere the state has been set to `RequestState.Finished` (which is not being set yet) or if the contract was started (`RequestState.Started`) and the contract duration has lapsed.
To facilitate `isFinished`, the contract start time (`startedAt`) was added to calculate the contract end time.
Mark Spanbroek
Eric Mastro