Given the following code:

// @flow

flowIDEBroke('')

function flowIDEBroke(x: number): number {
  return x
}

With [email protected] there is no error message, but running flow from the terminal shows:

index.js:3
  3: flowIDEBroke('')
     ^^^^^^^^^^^^^^^^ function call
  3: flowIDEBroke('')
                  ^^ string. This type is incompatible with
  5: function flowIDEBroke(x: number): number {
                              ^^^^^^ number


Found 1 error

If you revert using

apm install [email protected]

the errors in the editor pop back up.

This gives very accurate jump to definition, even across files. FINALLY

Leads to a lot of accidental completions when populating an object's properties:

{
  x: 1,
  y: 2, // <---- autocompletes when I press enter for a new line
}

https://gfycat.com/RewardingTameIberiannase

Hey first of all, I love this project. Thanks for putting in the hard work.

I recently updated and noticed that this project now supports code coverage which is awesome. Unfortunately, I had to turn it off because it gets a bit distracting with everything being underlined.

I'd like to suggest an alternative format and what https://github.com/joefitzgerald/go-plus does:

In this format, the uncovered code is in the columns.

I would like to use utiltiy type $Diff (https://flow.org/en/docs/types/utilities/#toc-diff)

but it shows $Diff is not definied

Is there a way to set a path to the flow bin you want to use? I was trying to contribute to flow, so I would need to use the compiled, built flow binary instead of using one that I npm install. Is there a way I can set the compiled flow binary that is in some other directory to be the one used by flow-ide? I tried searching the settings for the package, although I don't think there is any configuration available there.

To avoid stuff like this:

In a larger projects flow tend to take time to recheck printing 'flow is rechecking' message before it's finished. In such projects flow-ide is pretty much unusable resulting in this:

I would expect it to handle such cases gracefully.

In atom, the only displayed error message is the one on line 50. The only way to get to the useful information is to run flow in the terminal, and to see the full output, which gives me more clues about the issue, namely the "error" type issue.

src/auth/auth.reducer.js:50
 50:       return {
                  ^ object literal. This type is incompatible with the expected return type of
 34: export default function auth (state: State = initializeState(), action: Action):State {
                                                                                     ^^^^^ object type
  Property `error` is incompatible:
     26:   error: null
                  ^^^^ null. This type is incompatible with
     11:   error: Object
                  ^^^^^^ object type

Is there a way to show full output? (For the 26 and 11 line to be highlighted in addtion / instead of the 50 one). Seems to me that the errors on 26 / 11 are more useful.

After upgrading the package linter to 2.1.0 and flow-ide to 1.4.0, I started receiving the following error:[Linter] Error running Flow IDE Error: Process execution timed out(…)

Is there anything special to get this work in atom? I tried it out-of-the-box and also tried overwriting the executable location in the settings.

Given: https://flow.org/try/#0PTAE...

This is the output in atom:

This is the output of flow on the command line

person.js:12
 12:     return propEqual(p1, p2, "name");
                ^^^^^^^^^^^^^^^^^^^^^^^^^ function call
  8:     return props[prop] === otherProps[prop];
                                ^^^^^^^^^^^^^^^^ access of computed property/element. Computed property/element cannot be accessed on possibly null value
  8:     return props[prop] === otherProps[prop];
                                ^^^^^^^^^^ null

person.js:12
 12:     return propEqual(p1, p2, "name");
                ^^^^^^^^^^^^^^^^^^^^^^^^^ function call
  8:     return props[prop] === otherProps[prop];
                                ^^^^^^^^^^^^^^^^ access of computed property/element. Computed property/element cannot be accessed on possibly undefined value
  8:     return props[prop] === otherProps[prop];
                                ^^^^^^^^^^ undefined


Found 2 errors

So the output of flow includes one additional line (12: ...) that is actually really useful as it points to where the function is called. In a project where this is a helper function that lives somewhere in a helper.js this gets a little bit harder to track down the actual errors.

The additional line is also returned in the JSON that is available here: https://github.com/steelbrain/flow-ide/blob/v1.4.2/lib/helpers.js#L38-L54. It's stored on error.operation.

Specs:

• Windows 10.
• Atom 1.19.0-beta3
• Hyperclick 0.1.2
• flow-ide 1.6.0

Description

Everytime I try to jump to a definition, this happens:

And my PC freezes. The WMI Provider Host and those bunch of nodejs instances only appear when I try to jump to a definition.

While you try to stabilize this functionality, can you add an option to disable it? Right now I cannot use flow-ide because of that (or I need to be careful not to try to jump to a definition.)

Thank you for your work :)

I regularly get the following errors in the dev console while writing flow code:

Uncaught Error: write EPIPE
  exports._errnoException @ util.js:1026
  afterWrite @ net.js:795

[exec] Process exited with no-zero code
  (anonymous function) @ index.js:58 (= ~/.atom/packages/flow-ide/node_modules/sb-exec/lib/index.js)
  emitTwo @ events.js:106
  emit @ events.js:191
  maybeClose @ internal/child_process.js:877
  (anonymous function) @ internal/child_process.js:334
  emitOne @ events.js:96
  emit @ events.js:188
  _handle.close @ net.js:493

Sometimes I only get the Process exited but most of the time I get both errors in the above order.
It seems that neither of them causes any problems while writing flow code.
Maybe this is not a big problem and maybe even "expected", but it feels wrong to pollute the dev console with errors whereas they are not really errors.

Versions:

MacOSX 10.11.6

Atom    : 1.16.0
Electron: 1.3.13
Chrome  : 52.0.2743.82
Node    : 6.5.0

apm  1.16.1
npm  3.10.5
node 6.9.5 x64
python 2.7.10
git 2.10.1

flow-ide 1.4.2
flow 0.44.1

flow-ide settings:
executable path = ""
only active when .flowconfig exists [x]

As of today (after some fairy heavy refactoring) inline linter errors stopped showing up. I still see the warnings in the issues shown at the bottom of the window and I can confirm that these issues match the issues found when manually running flow and they do update if I intentionally introduce new errors.

I still see linter-eslint errors inline so I don't think the issue is affecting all of my linter implementations.

On version 1.1.5, tried 1.1.6 but nothing worked then

Flow 0.31 if it matters. Flow server is running if that also matters

I see there is another plugin trying to accomplish the same thing.

https://github.com/lukehoban/atom-ide-flow

Do you think it is worth approaching the author to combine your efforts into a single repo. Shared efforts and merged communities. x

Hi I was using this plugin but when using non-relative paths as disscussed in here http://moduscreate.com/es6-es2015-import-no-relative-path-webpack/, the flow-ide can't seem to find the module. (i.e. throws a warning required module not found)

We could use Atom APIs to get word boundary at a range and by default show it as clickable, and after the user clicks it we can invoke Flow to jump for us

Fixes #103

Thoughts @leos @lloiser?

This is a tricky one as I'm not able to reliably reproduce the problem.

Sometimes, Atom simply closes without any warnings or errors. "Crash" might not even be the right word, it looks more like a clean shutdown. The only problem is that I didn't initiate it or expect it to happen :)

While I can't be 100% certain it's caused by flow-ide, here are some observations:

• It started happening after I upgraded flow-ide to 1.4.2
• It happened for a day or two, until I disabled the flow-ide package.
• I kept the package disabled for several weeks and had none of these crashes in that time.
• Recently, I upgraded flow-ide to 1.5.0 and re-enabled it.
• After this, the crashes started happening again.
• Soon after that, I disabled the package again and haven't had any crashes since.

Do you have any ideas what to check for (some kind of logs maybe?) or how I might be able to reproduce it?
I'm running Atom on Windows 7.

Greetings,

First and foremost nice work on this project.
I am curious if it is possible to show jsdoc style documentation in the suggestions ?
I'm not sure if this would be a flow-ide or flow type of configuration.

Thank you for your time,

Ryan

Every now and then flow-ide puts a lot of strain on my system and I see a ton of flow autocomplete processes in Task Manager.

I haven't confirmed this, but my current theory is that when Flow is performing some large computations (sometimes it takes over a minute on my laptop), every autocomplete process will have to wait until this computation is complete. But every new keypress in the editor spawns a new autocomplete process. So this results in a large number of processes which slow down the computer, which probably makes the large Flow computation take even more time, starting a feedback loop of slowness.

If this theory is correct, perhaps a good solution would be to limit the number of autocomplete processes that are spawned? e.g. if there are, say, 5 already running, don't start any new ones until those have completed.

If we write this:

export function multiply(a, b) {
  return a*b;
}

Then flow will complain that a and b aren't annotated. The reason is explained here (basically: "Flow is stricter for any exported modules in order to avoid hard-to-find issues or bugs across different files.").

It'd be cool if I could use this flow plugin just for type inference (without annotation - basically just for smarter linting), but because of this thing with export, it doesn't look like it's possible.

Have I missed anything? Doesn't look like there's an option to suppress this complaint, but I am new to flow.

I figured I'd ask this question here and not in facebook/flow because I doubt they would care about using flow just as a "smarter linter" like this - but maybe I am wrong, so of course feel free to close if it's not relevant to this project.

Thanks!

Just like the title says. Making an issue so I don't forget

It's possible to do jsx props autocomplete ?

// @flow
function A(props: {a: string}) {
    return <div/>
}
function B() {
    return <A/> // Ctrl+Enter does not autocomplete A props
}

Hello, here is the type definition:

This is the autocomplete result:

The expected behavior is not to have that underscore as parameter.

Thank you for you work :)

This plugin runs Flow in the directory containing the current file, rather than the root project directory Atom is open to, or the project root as defined by the first parent directory with a .flowconfig file. This creates a problem when you have a Flow config that depends on <PROJECT_ROOT> in order to resolve certain modules with module.name_mapper, causing errors that only occur in flow-ide but not when running flow from the project root directory.

Running this package I get a noticeable hit on Atom's performace, with the editor sometimes halting for a couple seconds. Once I disable it I get the expected performance out of Atom. How can I get to the source of this issue? I really love flow-ide and want to keep using it.

Any time I post an issue about Windows in a flow related project, I can picture the maintainer going "ugh...who the hell uses Windows", but I'll post it anyway, haha 😼

So, on my macbook I've been using this plugin and it works as expected.
On Windows though, it works perfectly for reporting errors, and the autocomplete information looks accurate, except that there's no types shown in the autocomplete (the autocomplete looks like it does in vanilla Atom).

Any thought on what could be causing this or where I could start debugging it?

Argument to path.dirname must be a string

Object.dirname (/Applications/Atom.app/Contents/Resources/app.asar/src/electron-shims.js:9:10)
Object.<anonymous> (/Users/samhavens/.atom/packages/flow-ide/lib/index.js:168:43)
undefined.next (null:null:null)
step (/Users/samhavens/.atom/packages/flow-ide/lib/index.js:7:273)
<unknown> (/Users/samhavens/.atom/packages/flow-ide/lib/index.js:7:443)
updateCoverage (/Users/samhavens/.atom/packages/flow-ide/lib/index.js:7:99)

Atom 1.8.0, Linux 64. Just installed the package, opened a file with valid flow annotations, got this:

Don't much know what to do next :)

Can I use this together with Nuclide? (for other nuclide functionalities).

I want to see code not covered highlighted with yellow. Is it currently possible with this plugin? If no - it would be very useful to have

I tried to build a Windows version of Flow.
It works, but throws something to stderr but I cannot figure out why.
Due to this, flow-ide shows an error and interrupted linting.

I tried to set throwOnStdErr to false in exec options, and it works OK.
The autocompletion seems to work with this option and ignoreExitCode: true.
The modified code is in my fork currently. But I have not tested it in other platforms.

Hey there, first off, awesome project. Thanks for the hard work.

Just started using flow-ide and for some reason with this code I'm getting 2 errors for both functions saying that its uncovered code.

type VisibleAction = {
  type: 'workspace/commandPalette/VISIBLE', payload: { visible: boolean }
}

export const hideCommandPalette = (): VisibleAction => ({
  type: 'workspace/commandPalette/VISIBLE',
  payload: { visible: false },
});

export const showCommandPalette = (): VisibleAction => ({
  type: 'workspace/commandPalette/VISIBLE',
  payload: { visible: true },
});

Released in flow v0.42.0. This command will allow establishing a connection to flow to avoid multiple spawns

I'm trying to setup this package, but I can't get it working and there doesn't seem to be documentation on how to get started. I installed the plugin in atom, opened my flow project (with a valid .flowconfig), and I'm not getting any flow feedback. Also tried adding ./node_modules/.bin/flow as Executable Path in Settings but no luck.

Can someone lay out the steps for setting up this package?

I think I found a bug with flow-ide where if you have types that are being created from other files nested more than 1 level deep flow-ide begins to read the values as any and flow itself says there are no errors. I can create an example app if you would like me to demonstrate.

[Enter steps to reproduce:]

1. ...
2. ...

Atom: 1.18.0 x64
Electron: 1.3.15
OS: Microsoft Windows 10 Enterprise
Thrown From: flow-ide package 1.5.0

Stack Trace

Failed to activate the flow-ide package

At Package not initialized

Error: Package not initialized
  at /packages/nuclide/modules/nuclide-commons-atom/createPackage.js:48:15)
  at Provider.module.exports.Provider.provide (<embedded>:115569:58)
  at ServiceHub.module.exports.ServiceHub.provide (<embedded>:68815:26)
  at Package.module.exports.Package.activateServices (<embedded>:69312:77)
  at Package.module.exports.Package.activateNow (<embedded>:69122:22)
  at <embedded>:69091:38
  at Package.module.exports.Package.measure (<embedded>:68965:21)
  at <embedded>:69084:32
  at Package.module.exports.Package.activate (<embedded>:69081:40)
  at PackageManager.module.exports.PackageManager.activatePackage (<embedded>:20537:40)
  at <embedded>:53409:35
  at exit (<embedded>:53018:22)
  at triggerExitCallback (<embedded>:115985:17)
  at ChildProcess.<anonymous> (<embedded>:116015:17)
  at emitTwo (events.js:106:13)
  at ChildProcess.emit (events.js:191:7)
  at Process.ChildProcess._handle.onexit (internal/child_process.js:215:12)

Commands

     -3:40.6.0 settings-view:open (input.hidden-input)
     -3:36.1.0 core:backspace (input.hidden-input)
     -1:25.9.0 core:confirm (input.hidden-input)

Non-Core Packages

atom-beautify 0.30.3 
auto-detect-indentation 1.3.0 
auto-indent 0.5.0 
autocomplete-modules 1.6.10 
duotone-dark-earth-syntax 2.1.0 
duotone-dark-forest-syntax 2.1.0 
duotone-dark-sea-syntax 2.1.0 
duotone-dark-space-syntax 2.1.0 
duotone-dark-syntax 2.1.0 
duotone-light-syntax 2.1.0 
editor-stats 0.17.0 
editorconfig 2.2.2 
emmet 2.4.3 
emmet-jsx-css-modules 1.0.0 
file-icons 2.1.8 
firewatch-syntax 1.1.1 
flow-ide 1.5.0 
git-plus 7.9.3 
haskell-grammar 0.4.0 
highlight-selected 0.13.1 
hyperclick 0.1.1 
language-babel 2.66.0 
language-graphql 0.9.0 
language-ini 1.19.0 
language-javascript-jsx 0.3.7 
language-lua 0.9.11 
language-ocaml 1.9.5 
language-postcss 1.3.1 
language-rust 0.4.11 
language-swift 0.5.0 
language-thrift 1.0.2 
linter 2.2.0 
MagicPython 1.0.10 
merge-conflicts 1.4.5 
minimap 4.28.2 
nuclide 0.236.0 
nuclide-format-js 0.0.36 
nuclide-test-runner-jest 0.3.0 
pristine-ui 1.0.15 
react-native-snippets 0.3.0 
regex-railroad-diagram 0.19.3 
set-syntax 0.3.2 
sort-lines 0.14.0 
tool-bar 1.1.0 
tree-view-git-branch 0.1.2 
tree-view-git-status 1.4.0 
tree-view-open-files 0.3.0 
wakatime 7.0.3 

I think it would be helpful for completions to automatically show on a target object after typing a dot:

user.

Let's assume I want to complete fullName. Currently, autocomplete doesn't show up until I type f.

Let's take a look on the first example from https://flow.org/en/docs/types/maybe/ or any other live example from this site.

Both places (type definition and arg value) are hightlighet with error. Cause Flow does not know if you make error in type definition or in value declaration. So it's better to highlight errors by all locations which flow provides in message[] (open JSON tab in REPL) for exact error.

Now one IDE for now (flow-ide, nuclide, someone vscode plugin) does not highlight error in all places.

The worst experience what I getting is when I working with React Components covered by flow. If I provide incorrect property to component, error highlighted in Component declaration (in other file). Not in that file in which was written incorrect property. An it is sad.

Brings here previous conversation from #94 (comment)

@nodkz

Can we improve this plugin by displaying errors in several locations (different files)?

I use Flow for React components, and when I miss required property for component (in SwipeMagicRenderer.js)

the error displays only in a file SwipeableResizer.js where the component is defined:

Flowtype provides the following information for this Error (full json output gist):

{
  message: [
    { loc: { source: '..../SwipeableResizer.js' } },
    { loc: undefined },
    { loc: { source: '..../SwipeMagicRenderer.js' } },
  ]
}

But in function toLinterReference it takes only first file. But for better developer experience will be better to display error in all locations:

function toLinterReference(messages) {
  for (let i = 1, length = messages.length; i < length; i++) {
    const message = messages[i]
    if (message.loc) {
      return {
        file: message.loc.source,
        position: locationToRange(message.loc)[0],
      }
    }
  }
  return null
}

PS. Similar issue for Nuclide I opened 1 year ago, but they did not understand me.

PSS flow-ide plugin is much-much better. And can be better even more. Thanks for such awesome thing!

@steelbrain

I had implemented what you're saying in my Atom Hack package, which's linting part was seperated and is now the linter package, I dropped it for plenty of reasons.

The first and most important one is how do you distinguish between a normal message and a part of any other message? Keep in mind that blue underlines are already part of Info messages, also what do we show there? All of the message or just the relevant part? If all, how's the user going to know which part is the current one.

The second most important reason is performance, every new added message will be searched for references for not just the current text editor but messages from all text editors and will be added/removed when the text editor is closed and opened up again so this scales pretty horribly.

Also the support for standardized traces was removed in Linter Messages v2 to allow for a markdown string so v2 providers are inherently incapable of such a thing

@lloiser

@nodkz I would agree with you if flow would provide a better error message for each entry in message. Unfortunately flow's errors often only make sense if you read the complete error which is now visible in the error description.
I will merge this now. If you still think your suggestion is valid please open another issue and/or a pull request.

Brain, Lukas I mean that whole error (combined from all message entries like it already did in #94) will be displayed by all locations (unique locations gathered from all entries). Like it flowtype.org does (see first two screenshoots). And yep, I'm ready to Atom performance degradation, in favor forgetting calling flow via terminal and exploring whole error and all locations there. BTW this feature (highlight error in all places) can be hidden by option for this atom plugin. In all my apps and OSS packages I have 0 flowtype errors, so with this feature I will not have perf degradation, but obtain better dev exp.

I tried a number of different versions of flow and the latest one that seem to work with the plugin is 0.37.0.

Flow installed via brew doesn't seem to work too. It always times out with this message: linter-registry.js [sm]:137 [Linter] Error running Flow IDE Error: Process execution timed out(…)

Activity Monitor shows a number of active "flow" processes.

Is there any restrictions on what version of flow to use? Am I missing something?

Why does Flow-IDE run on files that aren't flow typed? It's a huge waste.

Not sure if this is too big of a reach for flow, but it'd be nice to be able to hover over variables and see what flow thinks their types are.

Autocomplete doesn't work on functions imported like this: import { validateArgs } from './validate-args'.

However if I go to validate-args.js, the autocomplete works great. It knows the types, can autofill the arguments for me.

This is a problem with autocomplete-flow as well, so I came over to try flow-ide to see if it works. Is this a problem with flow itself? I'd really like to see this functionality, it would save me time! Perhaps I'm using the plugin incorrectly. Thanks!

I could be completely wrong but I have a small suspicion that flow servers are spawned and never deleted. My CPU runs like crazy after opening and closing a few atom projects.

Of course, atom has never been the best editor in terms of performance but kind of curious about this.

Otherwise this is really good haha. Favourite flow IDE.

Aside from the %age, a very nice feature of Nuclide is showing you with a highlight what code is covered.

https://nuclide.io/docs/languages/flow/#type-coverage

Can I suggest this go on the roadmap? 😅

I put // @flow at the top of all my files, so when it's on a file that looks like this:

// @flow
import Table from './Table';
import Table__Row from './Table__Row';
import Table__Cell from './Table__Cell';

export {
    Table,
    Table__Row,
    Table__Cell
};

the coverage reports NaN%

Right now the readme has one-liner comparisons to Nuclide, linter-flow, and autocomplete-flow. Any chance a comparison to ide-flow could be added? It seems like that's the most directly comparable, and if not, it's the one that easiest to confuse with this package by name.

Original discussion here: steelbrain/linter#1198 (comment)

Linting current file is useful is that when developing with live reload / hot reload enabled (say react-native). Saving file would trigger the app to reload and state to be lost. So saving a file just to check flow / lint errors (other than the extra keystroke) is not ideal because it causes interruption in development context and might require one to navigate in the UI back to the original state again.

This could be an option that user could choose to turn on / off.