-
Notifications
You must be signed in to change notification settings - Fork 790
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
rlp: v3 updates from integration #1648
Conversation
export type Input = string | number | bigint | Uint8Array | List | null | undefined | ||
export type Input = string | number | bigint | Uint8Array | Array<Input> | null | undefined | ||
|
||
// Use interface extension instead of type alias to | ||
// make circular declaration possible. | ||
export interface List extends Array<Input> {} | ||
export type NestedUint8Array = Array<Uint8Array | NestedUint8Array> | ||
|
||
export interface Decoded { | ||
data: Uint8Array | NestedUint8Array | ||
remainder: Uint8Array | ||
} | ||
|
||
export interface NestedUint8Array extends Array<Uint8Array | NestedUint8Array> {} | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I didn't see any type issues changing to circular declaration here, so perhaps TypeScript added support for this from when the comment was originally written.
export function arrToBufArr(arr: Uint8Array | NestedUint8Array): Buffer | NestedBufferArray | ||
export function arrToBufArr(arr: Uint8Array | NestedUint8Array): Buffer | NestedBufferArray { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm not quite sure why I needed to do this final line twice, I thought I was getting pretty good at understanding these overloads too hehe, maybe a TypeScript wiz can help inform me, but I believe I set it up correctly.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Actually, instead of doing this, it would be much cooler if the return type was the same shape array as the input type but Buffers replaced with Uint8Arrays. That would probably require much less casting afterward. I'm not sure if this can be accomplished with generics, but I'll read the documentation and try to see.
Codecov Report
Flags with carried forward coverage won't be shown. Click here to find out more. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hmm, can you please test this new typing of the conversion methods on the v3
PR on #1630? I just injected the typings in there and get a lot of "No overload matches this call." errors already in the Util
library when trying to compile.
Maybe these all-catching conversion methods aren't so practical when it comes to typing and one should rather consider 1-to-1 methods with a single input and output? 🤔
Otherwise one would need to always re-cast when these methods get applied.
I guess in most (all?) cases the structure of the input is clear, e.g. if something is a simple Buffer or a Buffer array?
const encodedAsBuffer = Buffer.from(encoded) | ||
const decoded: Uint8Array[] = RLP.decode(encoded) | ||
const decodedAsBuffers = arrToBufArr(decoded) | ||
``` |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just a note: when we release v3 we should add this to the README to for greater visibility (alternative we can also already add now with a "(not yet released)" addition after "v3" and then remove on v3 release.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
sounds good, I already updated the README but can add a section for Buffer compatibility since the topic isn't directly discussed there.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hmm, can you please test this new typing of the conversion methods on the v3 PR on #1630? I just injected the typings in there and get a lot of "No overload matches this call." errors already in the Util library when trying to compile.
Just tried it out, this was the one place that error'd out: (this is because the version on that branch uses any
so it basically skips type checks)
Type 'Decoded' is missing the following properties from type '(Uint8Array | NestedUint8Array)[]': length, pop, push, concat, and 32 more.
const values = arrToBufArr(RLP.decode(Uint8Array.from(serialized)))
This is because we don't know what type RLP.decode
outputs, so this can be fixed with:
const values = arrToBufArr(RLP.decode(Uint8Array.from(serialized)) as Uint8Array[])
then we have to cast as Buffer[]
for fromValuesArray(values: Buffer[])
because NestedBufferArray
!== Buffer[]
which I think is okay. So:
const values = arrToBufArr(RLP.decode(Uint8Array.from(serialized)) as Uint8Array[]) as Buffer[]
a little ugly but better than using any
?
My overload typings above just help if you pass in a plain non-array Uint8Array or Buffer you get the non-array version back, so there should be less need for manual overrides in those cases to cast.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
have added a Buffer compatibility section to the readme here: https://github.com/ethereumjs/ethereumjs-monorepo/pull/1648/files#diff-68eff8537817899e4ff13ddb30d65da860ae9de825c4426c5551b3bf5ede53a6R35 let me know if it looks good
|
||
// Use interface extension instead of type alias to | ||
// make circular declaration possible. | ||
export interface List extends Array<Input> {} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not sure if this removal is worth the risk of breaking something on this release, in case someone has imported and used? Maybe we just keep and remove along v3?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Well, v3 is a breaking release so the types are different and not backward compatible
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
But this is not for v3 but for v2.
So if you remove an existing interface you might break something for some people? 🤔
Or are we somewhat talking past each other here? We can also jump on a quick call later my, earlier your day, mostly available. 🙂
|
||
export interface Decoded { | ||
data: Uint8Array | NestedUint8Array | ||
remainder: Uint8Array | ||
} | ||
|
||
export interface NestedUint8Array extends Array<Uint8Array | NestedUint8Array> {} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not sure, might this interface
-> type
switch cause compatibility problems as well?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is a new type that didn't exist in v2
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I also have this in my current master
branch, how can this be then? 🤔
bf71865
to
47cad6b
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ok, I am confused here on various stuff, e.g. there is also an ES2020
target in here.
So to recap: My current assumption is that this PR is for an RLP v2.x
release which does some preparation for an upcoming v3 release. Is this correct or am I mixing things up?
const decodedAsBuffers = arrToBufArr(decoded) | ||
assert.deepEqual(bufferList, decodedAsBuffers) | ||
``` | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
👍
|
||
// Use interface extension instead of type alias to | ||
// make circular declaration possible. | ||
export interface List extends Array<Input> {} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
But this is not for v3 but for v2.
So if you remove an existing interface you might break something for some people? 🤔
Or are we somewhat talking past each other here? We can also jump on a quick call later my, earlier your day, mostly available. 🙂
|
||
export interface Decoded { | ||
data: Uint8Array | NestedUint8Array | ||
remainder: Uint8Array | ||
} | ||
|
||
export interface NestedUint8Array extends Array<Uint8Array | NestedUint8Array> {} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I also have this in my current master
branch, how can this be then? 🤔
Update: Ah, now I'll slowly get my misconception. We do have RLP v3 already on Wow. That was a long one. 🙄 😛 Hmm, I find this a pretty much unlucky decision though with the scope of BigInt integration not really decided. Cannot really recall why we had been so confident on this? 🤔 I would really very much suggest that we cherry-pick the RLP v3 commits out to a separate branch and then do the releases from there and revert on |
Any updates @ryanio ? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
* rlp updates * util: add arrToBufArr and bufArrToArr and tests * util updates, add deprecation notice for re-exports * rlp tsconfig: include rootDir for composite option * remove util exports deprecation notices * rlp: add readme note for buffer compatibility * undo capitalization
Sorry, don't want to delay here that much further, but maybe two questions:
|
@holgerd77 this is good enough for browsers. The main thing was "0 deps", and "no buffers". As for ESM, could I suggest to not add it to v2? People will upgrade to v3 sooner in this case 😄 |
Ok, I'll trust you on this one, but just for me to get a better picture (really do not have that much browser-need feeling TBH, would be nice if we expand a bit on this over time 😬): what then is the main browser advantage here, if you integrate this into a build chain (Webpack, whatever) wouldn't this be polyfilled anyhow? Is there any usability advantage here? Or are you expecting foremost some performance gains?
Hehe. I am not sure if you are underestimating build targets though. I would assume there are various lib users out there who are just not ready yet for a BigInt switch, so I assume that the v2 release will still be out there for some time (and that would be an argument to also bring generic improvements). Metamask e.g. has Chrome 66 in it's minimal release target, which is still one off from BigInt introduction (not to speak about other ES2020 features. So I wonder if we should really at least disallow some of the newer language constructs/methods per ESLint? 🤔). |
That doesn't always happen, not all bundlers auto-shim buffers. My main goal, as usual, is supply chain auditability - lack of bn.js makes the code much easier to reason about. As for MM not supporting bigints - well, they'll need to bump their build target from the version which has less than 0.7% of global usage. Otherwise, they won't be able to use our new js-ethereum-cryptography. Old appe and old build targets can continue to use rlp v2. No one shuts them down, after all. Just want to make transition to bigints faster, because I know how hard it can be to ask folks to upgrade. |
Ok, I've now published the new version We'll keep the release a bit under radar until things are a bit tested and there is some first feedback. |
* VM: Adopted Backwards-compatible Dynamic Gas Cost Refactoring (#1553) * VM: report dynamic gas values in `fee` field of `step` event (#1364) * vm: codes: add dynamicGas property vm: codes: make codes more typesafe * vm: create no-op dynamic gas handlers * vm: first batch of dynamic gas up to 0x3f * vm: add other opcodes to gas map vm: change step event fee type from number to BN vm: deduct dynamic gas vm: fix stack peek vm: do not add gasLimit to gas cost for CREATE * vm: move errors to gas map * vm: fix memory dynamic gas bug * vm: fix gas bugs caused by not considering base fee * vm: fix message call gas related bugs, clone current gas left * add typedoc for peek use underscore for unused peek params (and fix eslint config) format some comment newlines for readability switch from require to import for exceptions * simplify the 2929 state manager castings in runTx * add changelog entry * vm: add EIP1283 tests * vm: split non-eip2929 and eip2929 gas costs * vm: fix gas costs * vm: add early-hardfork coverage * vm: clarify pre-Constantinople SSTORE gas vm: clarify EIP-150 comment * run coverage for all state and blockchain tests, remove redundant istanbul run * vm: fix CALLCODE gas vm: explicitly clone gasLimit for maxCallGas * vm: remove TODO in interpreter * update defaultCost to BN, cast 2929 statemanager to simplify use syntax * use underscore for unused variables, simplify types since they can be inferred * vm: fix browser tests + fix rebase * VM: moved dynamic fee to dedicated dynamicFee field in step event and retained fee behavior and type for backwards compatibility * VM: aligned InterpreterStep and step event object property documentation, completed missing step event properties * VM: test fix * vm: fix hardhat e2e tests * vm: fix MSTORE opcodes * vm: added dynamicGas property to push0 (EIP-3855) opcode * hardhat e2e: add temporary workaround for skipping tests with inconsistent memory field * nit style: use underscore instead of comment out unused variable Co-authored-by: Jochem Brouwer <jochembrouwer96@gmail.com> Co-authored-by: Ryan Ghods <ryan@ryanio.com> * ci: bump karma-typescript version (#1631) * bump karma-typescript to new release * bump libp2p-bootstrap * update package-lock * VM: Jump dest analysis enhancements (#1629) * Change valid jumps to single array * Switch array to Uint8Array * use correct opt code * Check for jumpsub * Address feedback * Address fixes * More efficiency * Adjust if/else logic for efficiency * Move comments Co-authored-by: Holger Drewes <Holger.Drewes@gmail.com> Co-authored-by: Ryan Ghods <ryan@ryanio.com> * rlp: add to root readme (#1642) * add rlp to root readme * add carryforward for rlp flag since it is not always tested on every build unless its paths are modified * Client: Add tests to verify shutdown (#1641) * Add tests to verify client shutdown * Add libp2p test * Address feedback * most libp2p into separate file * VM: Add comparison testing between branchs for State Test Runner (#1634) * Add diff tester and performance * update script * Simplify script * Readme updates * move start * Update stashing logic in script * tx: input value testing (#1620) * tx: helper methods to generate tx values * tx: helper methods to generate tx values * chore: scaffold test and add extra to values * chore: add missing spread operator * chore: single value for each tx key * tx: add generateCombinations method and legacy + eip 1559 test cases * chore: remove console log * chore: use london hardfork for eip1559 * tx: refactor generateCombinations interface and properly compare hash * tx: deterministically randomly sample 1000 elements from array for testing * chore: remove v * chore: remove todo * chore: clearer variable declarations and types * tx: fix and simplify random sampling * tx: display tx hash in testing output * tx: convert buffer to hex for output log * ci: fix node-versions run for node <16 (#1653) * re-add updating to npm v7 for node versions <16 * only upgrade npm for node v <16 * fix bin/rlp js: node 12 doesn't support ES11 which added support for nullish coalescing operator (??) so we'll use ternary here alternatively we could write this file in TS and compile to e.g. dist/bin/rlp (like we do in the client bin), but maybe if the file gets more complicated, in its current state i don't think it's so neccessary * use same errorMsg format for JSON.parse, remove unneeded extra Uint8Array.from (already is uint8array) * rlp: v3 updates from integration (#1648) * rlp updates * util: add arrToBufArr and bufArrToArr and tests * util updates, add deprecation notice for re-exports * rlp tsconfig: include rootDir for composite option * remove util exports deprecation notices * rlp: add readme note for buffer compatibility * undo capitalization * Devp2p, Client: Fix duplicated debug messages (#1643) * devp2p: update debug package * blockchain: update debug package * client: update debug package * vm: update debug package * devp2p/dpt fix for #1485 * devp2p/eth: Fix for 485 * devp2p: Fix for #1485: add base debug * devp2p/dpt #1485: change to debug.extend() * devp2p:dpt:server change to degub.extend() * devp2p:eth change to debug.extend() * devp2p/les: change to debug.extend() * devp2p:rlpx: change to debug.extend() * rlps: change to debug.extend() * Update debug to use IP as first tag * vm, client: removed ProofStateManager interface, added optional getProof, verifyProof methods to StateManager interface (#1660) * Monorepo (& Tx, VM): Examples scripts in CI (#1658) * chore(examples): examples added to ci * chore(examples-ci): remove script from VM (for now) & rename examples workflow file * chore(ci): new script formwatted with prettier & example workflow changes to run with non-test branches Co-authored-by: Holger Drewes <Holger.Drewes@gmail.com> Co-authored-by: Holger Drewes <Holger.Drewes@gmail.com> Co-authored-by: acolytec3 <17355484+acolytec3@users.noreply.github.com> Co-authored-by: Gabriel Rocheleau <contact@rockwaterweb.com> Co-authored-by: ScottyPoi <66335769+ScottyPoi@users.noreply.github.com> Co-authored-by: Cesar Brazon <cesarbrazon10@gmail.com>
This PR picks updates from #1630 to merge onto master to prep the rlp v3 release.
Then #1630 can be rebased and retargeted to the
develop
branch to be included in the other packages over time as breaking releases add es2020 support for BigInt.Two new methods (with tests) added to util to help with
Buffer
<>UInt8Array
conversions:arrToBufArr
andbufArrToArr
, with usage examples added to the rlp changelog.