refactor: the async await extravaganza #811
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
lidel
commented
Nov 7, 2019
|
@alanshaw i don't think ipfs/js-ipfs#2591 landed in v0.40.0, it will land in next release, right? |
|
Yep! |
lidel
changed the title
lidel
force-pushed
the
refactor/dry-libp2p-config
branch
from
7132542
to
8d2ff1d
Compare
|
Need to switch to latest js-ipfs, so I am repurposing this PR for general async iterators refactor, context #843 This is WIP. |
lidel
force-pushed
the
refactor/dry-libp2p-config
branch
2 times, most recently
from
930e670
to
01c1925
Compare
lidel
added
P0
Switches JS API to async iterators where possible. Includes switch to libp2p config override via ipfs/js-ipfs#2591 BREAKING CHANGE: switched to Async Iterators version of JS API https://blog.ipfs.io/2020-02-01-async-await-refactor/
- the latest js-ipfs builds and runs in the browser - 4x delegates and preloads enabled - 2x webrtc star provides signaling Known Issues: - the node starts, but struggles to find content, and there is a separate issue of infinite restart when user enables and disables it too fast - import via quick import page fails due to API discrepancy
lidel
force-pushed
the
refactor/dry-libp2p-config
branch
from
3fadffa
to
4dda1b1
Compare
As usual, additional hackery was required to get hapijs work in Chrome Apps context. There seem to be an import bug in js-ipfs when using the latest ipfs-webui, so we stay at the old version for now (for Brave only)
modern js-ipfs is using it for timeouts
web-ext lint requires JS files to be under 4MB, and latest js-ipfs included more dependencies, so we moved some of them from "backgroundBundle" to "ipfs" one, making it under 4MB.
Chrome seems to be broken by additional serialization that occurs. Potential fix may be just getting API address and creating new instance of http-client in quick-import.html (that will remove cross-process boundary in Chromium)
In short, there is a discrepancy between how objects are passed between a page and background page within the same extension in Firefox and Chromium. Quick import in Chromium was broken, most likely by additional process boundary serialization that was not present in Firefox, could be related to upstream bug at: https://bugs.chromium.org/p/chromium/issues/detail?id=112163 Potential fix via trampoline from https://stackoverflow.com/a/21101597/11518426 is too brittle, as Chromium can break it in browser context at any time. It seems that ReadableStream produced by file.stream() is able to correctly pass the boundary between quick import page and the IPFS API object running in the background, so we use it with embedded js-ipfs, but for external node we use http client directly from quick upload page, and pass File object as-is, which takes 1/2 of time. Additional benchmarks can be found in #811
for now just embedded js-ipfs, but http-client will be fixed in one of future releases and will benefit from this as well
js-ipfs in brave does not work correctly and we can't update to the latest one because of other issues. embedded js-ipfs in brave goes away in next year, so we simply open imports on gateway when chrome apps runtime is used, which is good enough for experimental feature
embedded node in brave does not need it anymore, we reload entire extension on port change regular embedded never needed it
this enables webrtc-based signaling and delegates for experimental js-ipfs running in extension context (non-Brave) when no custom config is provided
lidel
force-pushed
the
refactor/dry-libp2p-config
branch
from
2236ecf
to
18ebc75
Compare
We were bundling TAR inside of Chromium package to make it work instantly when opened in Brave. Unfortunately it added 10MB to the package, which meant one-click install was slow in Brave, but also on every other Chromium browser. This replaces bundled TAR with prefetching it from a public gateway on the first run. The idea is that if user had access to Chrome Web Store to install extension, they have access to the gateway, so no need for bundling. We also precache regular webui in Chromium and Firefox via ipfs.refs
lidel
force-pushed
the
refactor/dry-libp2p-config
branch
from
18ebc75
to
ed319fa
Compare
lidel
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR:
TODO
quick import in Firefox
quick import in Chromium
refs: https://bugs.chromium.org/p/chromium/issues/detail?id=112163, potential fix via trampoline from https://stackoverflow.com/a/21101597/11518426, https://stackoverflow.com/a/24303394/11518426
ReadableStreamproduced byfile.stream()is able to correctly pass the boundary between quick import page and the IPFS API object running in the background.file.stream()+ bg-page http client → ~47sfile.stream()+ same-page http client → ~47sfile(Blob) + same-page http client → ~25s 💚file.stream()+ bg-page js-ipfs → ~1m40sfile.stream()+ bg-page http client → 💔 always errors after ~15sfile.stream()+ same-page http client → 💔 always errors after ~15sfile(Blob) + same-page http client → ~25s 💚file.stream()+ bg-page js-ipfs → ~1m30sfile(Blob) + same-page http client provide best performance for external node, whilefile.stream()+ bg-page js-ipfs make things work okay-ish when embedded js-ipfs is used (we don't expect people to import 4GB with embedded node so good for now)fix progress reporting
precache
context actions
import via context action -- seee Backwards compatibility of
ipfs.add(urlSource(url))js-ipfs#3195 firstembedded js-ipfs in Chrome
embedded js-ipfs in Brave
- [ ] switch window.ipfs to new API (or fill an issue to do it in separate PR)(window.ipfs has been disabled for a while)