Summary:
public
Remove the unused feature for async dependencies / bundle layouts. We can bring it back later, if needed.
Reviewed By: cpojer
Differential Revision: D2916543
fb-gh-sync-id: 3a3890f10d7d275a4cb9371a6e9cace601a82b2c
shipit-source-id: 3a3890f10d7d275a4cb9371a6e9cace601a82b2c
Summary:
The bundler class had duplicated code and parts that were hard to follow, because functions accepted heterogenous arguments, leading to a lot of conditionals.
This commit tries to remove duplication between build steps of different type of bundles, and by accepting less different types of arguments. These differences are now handled further up the call stack.
public
Reviewed By: sebmarkbage
Differential Revision: D2905807
fb-gh-sync-id: ef85ea0d461a9a06a4a64480e014a5324c4ef532
Summary:
node-haste shouldn't ever call process.exit and should leave it up to clients to shut down properly. This change just moves it out into the `Resolver` class – I'll leave it up to David and Martin to improve error handling for the rn packager :)
public
Reviewed By: davidaurelio
Differential Revision: D2889908
fb-gh-sync-id: 6f03162c44d89e268891ef71c8db784a6f2e081d
Summary:
Right now, a mock called `debug.js` shadows a node module called `debug`. When I made mocking more strict I didn't realize that it didn't include those mocks any more because requiring `debug` would result in a module like `debug/node.js` which doesn't have a mock associated with it. This diff changes it so it looks at the associated `package.json` to match the name up to mocks. This is the least invasive non-breaking change I can make right now. Yes, mocking strategies are basically fucked but I don't want the haste2 integration to have even more breaking changes right now. Consider this code to be temporary, I'll fix this and make the mocking system more sane mid-term.
public
Reviewed By: davidaurelio
Differential Revision: D2889198
fb-gh-sync-id: 58db852ed9acad830538245f7dc347365fe4de38
Summary:
Uses `fastfs.readWhile` to build the haste map
- cuts the time needed to build the haste map in half
- we don’t need to allocate memory for all JS files in the tree
- we only read in as much as necessary during startup
- we only read files completely that are part of the bundle
- we will be able to move the transform before dependency extraction
public
Reviewed By: martinbigio
Differential Revision: D2890933
fb-gh-sync-id: 5fef6b53458e8bc95d0251d0bcf16821581a3362
Summary:
The feature added in D2862850 only adds the module that requests shallow dependency resolution to the dependencies of a module. The reason why this is happens is because `collect` calls `response.addDependency` but if `recursive` is set to `false`, dependencies don't call `collect` and therefore don't get added to the resolution response. This fixes it by adding dependencies outside of the `collect` call.
public
Reviewed By: davidaurelio
Differential Revision: D2885152
fb-gh-sync-id: 8ab3b6c6b7cd45d59615a99ac87984a41b5d7025
Summary:
This syncs a couple of changes from GitHub that are necessary for jest:
* Expose `set` on Cache
* Add a function to `getAllModules` from the graph
* Explicitly dontMock `graceful-fs` in the fastfs test. node-haste2 has this manual mock which is still used in automocked tests (ugh!).
public
Reviewed By: davidaurelio
Differential Revision: D2885112
fb-gh-sync-id: b2060d5474ebee5aa13e6ba2bdf5879b46f3fd5f
Summary:
The FS as seen by an instance of a dependency graph is useful for clients of node-haste. This provides an accessor to fastfs so that the partial view of the filesystem can be queried. I was thinking of splitting out the fastfs creation to outside the constructor but the necessary refactoring doesn't seem worth it. The fastfs is intrinsically tied to the dependency graph and the `DependencyGraph` instance is meant to be a base-class for clients anyway. Both in react-packager and jest we are wrapping it with higher-level client specific containers, Resolver and HasteResolver respectively.
public
Reviewed By: davidaurelio
Differential Revision: D2885217
fb-gh-sync-id: 47a408b2682516bee9d6a4e6c61b9063817aaf22
Summary:
public
This fixes the problem that graceful-fs exposes an unpatched `close()` method, that will never trigger retries.
This behavior deadlocks the packager when trying to use graceful-fs’s `open()`/`read()`/`close()` methods.
See: isaacs/node-graceful-fs#56
Reviewed By: cpojer
Differential Revision: D2885512
fb-gh-sync-id: 71112d2488929bf1775fe9f8566fa03fd66b6bea
Summary:
Jest needs this for efficient caching of resolution responses.
public
Reviewed By: voideanvalue
Differential Revision: D2873291
fb-gh-sync-id: fee27d2ffdfe64bd68fdb4d9e4259e721b33631f
Summary:
public
At the moment, when the user changes a file we end up pulling the dependencies of the entry point to build the bundle. This could take a long time if the bundle is big. To avoid it, lets introduce a new parameter to `getDependencies` to be able to avoid processing the modules recursively and reuse the resolution responseto build the bundle.
Reviewed By: davidaurelio
Differential Revision: D2862850
fb-gh-sync-id: b8ae2b811a8ae9aec5612f9655d1c762671ce730
Summary:
public
Adds the ability to read files partially with `readWhile` to `Fastfs`
This feature will be used by for building the haste map, where we only need to read the doc block (if present) to extract the provided module name.
Reviewed By: martinbigio
Differential Revision: D2878093
fb-gh-sync-id: 219cf6d5962b89eeb42c728e176bf9fcf6a67e9c
Summary:
public
This adds an option to `Module` (and its callers) that allows to transform code before extracting dependencies. The transform function has to return a promise that resolves to an object with `code`, and optionally `dependencies` and/or `asyncDependencies` properties, if standard dependency extraction cannot be applied
Reviewed By: cpojer
Differential Revision: D2870437
fb-gh-sync-id: 806d24ba16b1693d838a3fa747d82be9dc6ccf00
Summary:
This reverts commit 888749220d.
The original commit didn't handle cases like .(ios|android|native).js, and vjeux is in favor of standardizing on .js instead of custom extensions like .jsx or .es6 everywhere.
> Unfortunately this pull request doesn't implement with .ios.jsx. But, when/if it does, it raises more questions like what happens if you have both ios.js and ios.jsx?
>
> Sorry to chime in super late but I actually think that this is harmful to support .jsx extension. We now live in a world where we have many transforms for es6, flow, jsx... Why would we add .jsx but not .flow or .es6. Supporting more extensions is only going to fragment tools even more.
https://github.com/facebook/react-native/pull/5233#discussion_r49682279
Closes https://github.com/facebook/react-native/pull/5296
Reviewed By: svcscm
Differential Revision: D2830784
Pulled By: bestander
fb-gh-sync-id: 9a7a4745803acbcbc5dba176b971c629c904bacd
Summary:
public
The reverted change doesn’t play nice with inline requires, let’s revert it for now.
I will bring it back after fixing it or adapting inline requires
Reviewed By: martinbigio
Differential Revision: D2854771
fb-gh-sync-id: 32fdbf8ad51240a9075b26502decb6328eed4b29
Summary:
public
Since the combination of node and haste modules (and modules that can be required as both node and haste module) can lead to situations where it’s impossible to decide an unambiguous module identifier, this diff switches all module ids to integers. Each integer maps to an absolute path to a JS file on disk.
We also had a problem, where haste modules outside and inside node_modules could end up with the same module identifier.
This problem has not manifested yet, because the last definition of a module wins. It becomes a problem when writing file-based unbundle modules to disk: the same file might be written to concurrently, leading to invalid code.
Using indexed modules will also help indexed file unbundles, as we can encode module IDs as integers rather than scanning string IDs.
Reviewed By: martinbigio
Differential Revision: D2842418
fb-gh-sync-id: 97addd28e964ac5f2b5081dcd3f36124d2864df8
Summary:
There's a long standing issue on open source (https://github.com/facebook/react-native/issues/4968). Until someone gets some free bandwidth to fix it lets at the very least improve the error message to guide users to the issue and suggest workarounds.
public
Reviewed By: mkonicek
Differential Revision: D2849051
fb-gh-sync-id: ef7913442ceabcab2076141bd13ab1ceeb529759
Summary:
We used to only cache the `dependencies` and `name` of a Module but we were actually accessing two more fields (async dependencies and the `isHaste` field) which meant we were always reading every single file from disk. In D2644383 I noticed that but realized that the Promise was cached, meaning we would read every file once *per instance* and I didn't think about cross-instance reads over time. My initial version added more caching (but also missed the `isHaste` field) but then I got rid of the cache call for `dependencies`. So my change from before didn't make anything worse but it also didn't make anything better. This change now caches everything until the contents of the file actually changes.
public
Reviewed By: martinbigio
Differential Revision: D2831569
fb-gh-sync-id: 74081abc0ce3ca96b4e56c3c9b6d24aa84f7496c
Summary:
It's possible that a mock doesn't have an associated real module that it maps to. This is actually very common in www, where we have JS mocks for dynamic PHP JS modules. The implementation I chose seems like the smartest one for now: if a module cannot be resolved, we look up whether we have a mock with the same id. If we do, we just resolve it. That's it! And it also only does the minimum amount of resolution necessary.
public
Reviewed By: martinbigio
Differential Revision: D2822277
fb-gh-sync-id: 7c9fbb6f69a0c0c85157c0650f5719d94a02410e
Summary:
I'm working on deploying haste2 with jest. This updates all the files that require changes for this to work and they are backwards compatible with the current version of jest.
* package.json was just outdated. I think haste1's liberal handling with collisions made this a "non-issue"
* env.js didn't properly set up ErrorUtils, also unsure why that isn't a problem in jest right now already?
* some things were mocking things they shouldn't
* Because of the regex that matches against providesModule and System.import, it isn't possible to list module names more than once. We have multiple tests reusing the same providesModule ids and using System.import with modules that only exist virtually within that test. Splitting up the strings makes the regexes work (we do the same kind of splitting on www sometimes if we need to) and using different providesModule names in different test files fixes the problem. I think the BundlesLayoutIntegration-test is going to be deleted, so this doesn't even matter.
public
Reviewed By: voideanvalue
Differential Revision: D2809681
fb-gh-sync-id: 8fe6ed8b5a1be28ba141e9001de143e502693281
Summary:
public
We're not planning to accept file removals in the short term on the HMR interface so lets bail when a file is removed (before this this we were throwing when trying to get the shallow dependencies).
Reviewed By: yungsters
Differential Revision: D2810534
fb-gh-sync-id: f2733382f4a2619e22bdf1163aa4180694fff9f8
Summary:
public
Before this diff we were only accepting the module that was modified but the user. This works fine as long as the user doesn't modify the dependencies a module has but once he starts doing so the HMR runtime may fail when updating modules' code because they might might a few dependencies. For instance, if the user changes the `src` a `Image` has to reference an image (using the new asset system) that wasn't on the original bundle the user will get a red box. This diff addresses this by diffing the modules the app currently has with the new ones it should have and including all of them on the HMR update. Note this diffing is only done when the we realize the module that was modified changed it's dependencies so there's no additional overhead on this change.
Reviewed By: vjeux
Differential Revision: D2796325
fb-gh-sync-id: cac95f2e995310634c221bbbb09d9f3e7bc03e8d
Summary:
We don't (yet) treat these the same as any other modules because we still have special resolution rules for them in the packager allowing the use of `providesModule`, but I believe this allows people to use npm react in their RN projects and not have duplicate copies of React. Fixesfacebook/react-native#2985.
This relies on fbjs 0.6, which includes `.flow` files alongside the `.js` files to allow them to be typechecked without additional configuration. This also uses react 0.14.5, which shims a couple of files (as `.native.js`) to avoid DOM-specific bits. Once we fix these in React, we will use the same code on web and native. Hopefully we can also remove the packager support I'm adding here for `.native.js`.
This diff is not the desired end state for us – ideally the packager would know nothing of react or fbjs, and we'll get there eventually by not relying on `providesModule` in order to load react and fbjs modules. (fbjs change posted here but not merged yet: https://github.com/facebook/fbjs/pull/84.)
This should also allow relay to work seamlessly with RN, but I haven't verified this.
public
Reviewed By: sebmarkbage
Differential Revision: D2786197
fb-gh-sync-id: ff50f28445e949edc9501f4b599df7970813870d
Summary:
public
Compute the dependencies of the bundle entry file just before sending HMR updates. In case the file that was changed doesn't belong to the bundle bail.
Reviewed By: vjeux
Differential Revision: D2793736
fb-gh-sync-id: f858e71b0dd5fe4f5b2307a22c6cef627eb66a22
Summary:
There's quite a bit of code scattered around the packager regarding ignoring the `providesModule` Haste pragma in any file that isn't in `react-native`, `react-tools` or `parse`. There is even a (passing) test case.
However, there's an edge case.
Take, for example, `fbjs`. It has a module inside of it called `ErrorUtils`. `react-relay` requires this file normally, in Common.JS style, by doing `require('fbjs/libs/ErrorUtils')`. But when `react-native` attempts to require `ErrorUtils` using the HasteModule format (in it's JavaScript initialization), it resolves the `fbjs` `ErrorUtils` module, instead of RN's `ErrorUtils`.
This happens, it turns out, because when a module is read (in `Module._read`), it's not caring about whether or not it should pay attention to `providesModule`, and is just assigning the `providesModule` value as the id of the module no matter what. Then when `Module.getName` is called, it will always use that `data.id` that was set, thus creating the wrong dependency tree.
This
Closes https://github.com/facebook/react-native/pull/3625
Reviewed By: svcscm
Differential Revision: D2632317
Pulled By: vjeux
fb-gh-sync-id: efd8066eaf6f18fcf79698beab36cab90bf5cd6d
Summary:
Here are some small fixes for issues we've encountered with very large RN projects (mostly huge dependency trees in `node_modules`).
cc amasad martinbigio
Closes https://github.com/facebook/react-native/pull/4880
Reviewed By: svcscm
Differential Revision: D2782834
Pulled By: mkonicek
fb-gh-sync-id: e316a62b84ba796b80ac819431414ebf27f7b566
Summary:
in ```fastfs.js ``` when ```getFile()``` got a exception it will print ``` Unable to find file with path: null ``` in terminal .
It's confused
Closes https://github.com/facebook/react-native/pull/4737
Reviewed By: svcscm
Differential Revision: D2752888
Pulled By: androidtrunkagent
fb-gh-sync-id: a366da1eea27c691248dcb17019f4462a639ea70
