Support singleton webpacker (3 and above)

[BookOfGreg]

Fixes#776
Fixes#775
Fixes#778
Fixes#780
Fixes#781
Fixes#782

Once again I could probably use some assistance with the appraisal gem but other than that should be OK.

Edit: This turned into a BIG refactor of the tests to split out support and I'll be honest my code isn't the tidiest. Still plenty of refactoring to go as I'm not sure I'm thrilled about how I've switched method definitions of Webpacker version numbers.

[AnatoliiD]

should be

Webpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configuration

[BookOfGreg]

Aaw whoops. Wondered why my test didn't pass XD Thanks

[BookOfGreg]

I've still got issues around Webpacker::Manifest.load, seems to be how I detect respond_to? may be wrong for that one.

[bradleesand]

The install generator won't work because entry_path was removed from webpacker config.

[bradleesand]

The new singleton implementation of Webpacker::Configuration no longer has entry_path. Instead, Webpacker::Configuration.source_path.join(Webpacker::Configuration.entry_path) becomes Webpacker.config.source_entry_path

[vikash-k-singh]

I have suggested these changes in issue #778 .Please have a look into it. I was trying to install react with rails and failed to run react installer command. I look into gem and found that method calling is not proper.

I fix this with minor changes in lib/generators/react/install_generator.rb file.

source_path is a instance method and was being called by class directly. Also entry_path was changed to source_entry_path in webpacker config. I have suggested these changes in #778

The main purpose for issue was #778 was to elevate exact issue and its solution along with it, as there was no fix until that time.

[basicBrogrammer]

Using webpacker 3.0.1 Webpacker.config.source_entry_path worked for me 👍

[vikash-k-singh]

It looks good, only a minor change suggested that webpack_configuration.entry_path should be changed to webpack_configuration.source_entry_path, as entry_path is renamed as source_entry_path in webpacker config file. File location for this file is lib/generators/react/install_generator.rb under method name javascript_dir.

[BookOfGreg]

I took on the code from #778 , thank you for that one! Makes it a little easier to update.

Currently working on trying to make the new config/webpacker.yml and it's env files play nicely with webpacker V1's paths.yml. It's putting my manifest in the webpacker.yml's output location with the paths.yml's manifest name so the webpacker helpers can't find the manifest.

Will take me some time to figure out all the subtleties of this before I'm done. Very welcome for someone to help with this!

Edit: Is there a nice way of changing the config files in test/dummy?

[vikash-k-singh]

In file lib/react/server_rendering/webpacker_manifest_container.rb, I am looking at a method name lookup_path at line no. 24, but I don't think it is defined anywhere in the both react-rails & webpacker gem. This will through an error once find_asset(logical_path) method is being called.

Please look into it! Are you guys loading any file with such variable?

Apart for this, hello @BookOfGreg, what kind of changes you wanna in test/dummy config files. Does it contains any Json data? Please share, hopefully I may change that!

[BookOfGreg]

Webpacker 1 and 2 have lookup_path, Webpacker 3 does not. and I'll have to compose a method of finding the actual path of a file.

[basicBrogrammer]

I'm not sure if this is related, but the component generator is putting my components in app/javascript/packs/components. To get this to work, I had to change the require context from 'components' to './components'

app/javascript/packs/application.js
var componentRequireContext = require.context('./components', true')

And also in my es6 components I have to require('../components/NameOfComponent')
Any thoughts?

[swrobel]

Seems like you might've messed up the indentation here

[bradleesand]

The <<- syntax allows the closing HEREDOC to be indented but it won't strip space from the content so this works just fine.

[willcosgrove]

I think this needs to be

defconfigWebpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configurationend

Because even in Webpacker 3, the Webpacker::Configuration constant is defined.

[tomwaddington]

Just ran into issue #778, and this fixed it for me. 👍

[BookOfGreg]

I'll admit that I'm at a bit of a dead end. On #779 I've got all of Travis passing except Rails 5 + Webpacker, but on that branch Webpack builds to public/packs but Webpacker looks in packs/packs instead. That is the case even if I peg Webpack at 1.15.x and Webpacker at ~> 1.0.
If a regular committer could help me out to get the tests passing on master then I could continue working here to get more Webpacker support.

We should really think about the maintainability of the Webpacker support as the manifest paths between versions seems really flakey and i'm not keen on the pattern of .responds_to? and defined? everywhere.

[BookOfGreg]

Master finally passes and this does mean Webpacker 3 support (And 1.1, 1.2, 2.0) on this branch.
I split out as much as I could do make it clearer but there is still a lot of mess I caused in the code, lots of refactoring to do.

[RiccardoMargiotta]

@BookOfGreg 😄

[rmosolgo]

Wow, this is really impressive! Thanks for all your work in digging in on this. It's nice to see that Webpacker has provided some APIs to support this kind of thing, too.

I see the green check, and the tests look good. So, what's still to go on this one?

[ArnonHongklay]

👍

[BookOfGreg]

This could go in to master it's current state as I believe it to be an improvement but frankly this could use some TLC and refactoring.

I'm certain all my MAJOR < 3 checks could be extracted into a class following the existing pattern.
There are probably better ways of integrating with the Webpacker.dev_server also.

I found some odd behaviour such as webpack 3 seems to read the wrong node_modules folder when loading in ../../../../react_ujs, so I used the actual npm package for it instead, that might cause some people in the wild issues that I can't even predict if they're messing with node context.

All these tweaks could be in future patches to this though.

Edit: Since I've modified the webpacker 1.1, 1.2 and 2 support, it would be good to see if it still works without bugs on an existing project for confirmation.

[Melonbwead]

You beast @BookOfGreg 👍 ❗️

[BookOfGreg]

I think we should probably merge this if we can be confident that it works. All existing tests pass. I can carry on with a tidy up beyond this.

Edit: apparently linting broke the tests. How odd. Will keep looking into it.

[BookOfGreg]

Tests fail on specific seeds:
bundle exec appraisal rails-5_no_sprockets_webpacker_3 rake test TESTOPTS="--seed=17880"

[BookOfGreg]

@rmosolgo If you like, and I'll keep PR'ing rubocops, test fixes and refactorings as patch-level changes after this. Be good to start getting some bug reports from folks so I know if this accidentally impacts anyone.

[rmosolgo]

That sounds great, let me go ahead with a release!

[gaiapunk]

I'm still having the #778 issue and was wondering if this latest merge had been release to fix it or if there is a workaround on my end that I can do before a fix is released, thanks for all the work you do!

[BookOfGreg]

@gaiapunk It's not been released as a gem yet but it's possible to use the master branch directly with:
gem 'react-rails', git: 'https://github.com/reactjs/react-rails.git', branch: 'master'

[BookOfGreg]

Fixes#776
Fixes#775
Fixes#778
Fixes#780
Fixes#781
Fixes#782

Once again I could probably use some assistance with the appraisal gem but other than that should be OK.

Edit: This turned into a BIG refactor of the tests to split out support and I'll be honest my code isn't the tidiest. Still plenty of refactoring to go as I'm not sure I'm thrilled about how I've switched method definitions of Webpacker version numbers.

[AnatoliiD]

should be

Webpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configuration

[BookOfGreg]

Aaw whoops. Wondered why my test didn't pass XD Thanks

[BookOfGreg]

I've still got issues around Webpacker::Manifest.load, seems to be how I detect respond_to? may be wrong for that one.

[bradleesand]

The install generator won't work because entry_path was removed from webpacker config.

[bradleesand]

The new singleton implementation of Webpacker::Configuration no longer has entry_path. Instead, Webpacker::Configuration.source_path.join(Webpacker::Configuration.entry_path) becomes Webpacker.config.source_entry_path

[vikash-k-singh]

I have suggested these changes in issue #778 .Please have a look into it. I was trying to install react with rails and failed to run react installer command. I look into gem and found that method calling is not proper.

I fix this with minor changes in lib/generators/react/install_generator.rb file.

source_path is a instance method and was being called by class directly. Also entry_path was changed to source_entry_path in webpacker config. I have suggested these changes in #778

The main purpose for issue was #778 was to elevate exact issue and its solution along with it, as there was no fix until that time.

[basicBrogrammer]

Using webpacker 3.0.1 Webpacker.config.source_entry_path worked for me 👍

[vikash-k-singh]

It looks good, only a minor change suggested that webpack_configuration.entry_path should be changed to webpack_configuration.source_entry_path, as entry_path is renamed as source_entry_path in webpacker config file. File location for this file is lib/generators/react/install_generator.rb under method name javascript_dir.

[BookOfGreg]

I took on the code from #778 , thank you for that one! Makes it a little easier to update.

Currently working on trying to make the new config/webpacker.yml and it's env files play nicely with webpacker V1's paths.yml. It's putting my manifest in the webpacker.yml's output location with the paths.yml's manifest name so the webpacker helpers can't find the manifest.

Will take me some time to figure out all the subtleties of this before I'm done. Very welcome for someone to help with this!

Edit: Is there a nice way of changing the config files in test/dummy?

[vikash-k-singh]

In file lib/react/server_rendering/webpacker_manifest_container.rb, I am looking at a method name lookup_path at line no. 24, but I don't think it is defined anywhere in the both react-rails & webpacker gem. This will through an error once find_asset(logical_path) method is being called.

Please look into it! Are you guys loading any file with such variable?

Apart for this, hello @BookOfGreg, what kind of changes you wanna in test/dummy config files. Does it contains any Json data? Please share, hopefully I may change that!

[BookOfGreg]

Webpacker 1 and 2 have lookup_path, Webpacker 3 does not. and I'll have to compose a method of finding the actual path of a file.

[basicBrogrammer]

I'm not sure if this is related, but the component generator is putting my components in app/javascript/packs/components. To get this to work, I had to change the require context from 'components' to './components'

app/javascript/packs/application.js
var componentRequireContext = require.context('./components', true')

And also in my es6 components I have to require('../components/NameOfComponent')
Any thoughts?

[swrobel]

Seems like you might've messed up the indentation here

[bradleesand]

The <<- syntax allows the closing HEREDOC to be indented but it won't strip space from the content so this works just fine.

[willcosgrove]

I think this needs to be

defconfigWebpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configurationend

Because even in Webpacker 3, the Webpacker::Configuration constant is defined.

[tomwaddington]

Just ran into issue #778, and this fixed it for me. 👍

[BookOfGreg]

I'll admit that I'm at a bit of a dead end. On #779 I've got all of Travis passing except Rails 5 + Webpacker, but on that branch Webpack builds to public/packs but Webpacker looks in packs/packs instead. That is the case even if I peg Webpack at 1.15.x and Webpacker at ~> 1.0.
If a regular committer could help me out to get the tests passing on master then I could continue working here to get more Webpacker support.

We should really think about the maintainability of the Webpacker support as the manifest paths between versions seems really flakey and i'm not keen on the pattern of .responds_to? and defined? everywhere.

[BookOfGreg]

Master finally passes and this does mean Webpacker 3 support (And 1.1, 1.2, 2.0) on this branch.
I split out as much as I could do make it clearer but there is still a lot of mess I caused in the code, lots of refactoring to do.

[RiccardoMargiotta]

@BookOfGreg 😄

[rmosolgo]

Wow, this is really impressive! Thanks for all your work in digging in on this. It's nice to see that Webpacker has provided some APIs to support this kind of thing, too.

I see the green check, and the tests look good. So, what's still to go on this one?

[ArnonHongklay]

👍

[BookOfGreg]

This could go in to master it's current state as I believe it to be an improvement but frankly this could use some TLC and refactoring.

I'm certain all my MAJOR < 3 checks could be extracted into a class following the existing pattern.
There are probably better ways of integrating with the Webpacker.dev_server also.

I found some odd behaviour such as webpack 3 seems to read the wrong node_modules folder when loading in ../../../../react_ujs, so I used the actual npm package for it instead, that might cause some people in the wild issues that I can't even predict if they're messing with node context.

All these tweaks could be in future patches to this though.

Edit: Since I've modified the webpacker 1.1, 1.2 and 2 support, it would be good to see if it still works without bugs on an existing project for confirmation.

[Melonbwead]

You beast @BookOfGreg 👍 ❗️

[BookOfGreg]

I think we should probably merge this if we can be confident that it works. All existing tests pass. I can carry on with a tidy up beyond this.

Edit: apparently linting broke the tests. How odd. Will keep looking into it.

[BookOfGreg]

Tests fail on specific seeds:
bundle exec appraisal rails-5_no_sprockets_webpacker_3 rake test TESTOPTS="--seed=17880"

[BookOfGreg]

@rmosolgo If you like, and I'll keep PR'ing rubocops, test fixes and refactorings as patch-level changes after this. Be good to start getting some bug reports from folks so I know if this accidentally impacts anyone.

[rmosolgo]

That sounds great, let me go ahead with a release!

[gaiapunk]

I'm still having the #778 issue and was wondering if this latest merge had been release to fix it or if there is a workaround on my end that I can do before a fix is released, thanks for all the work you do!

[BookOfGreg]

@gaiapunk It's not been released as a gem yet but it's possible to use the master branch directly with:
gem 'react-rails', git: 'https://github.com/reactjs/react-rails.git', branch: 'master'

[BookOfGreg]

Fixes#776
Fixes#775
Fixes#778
Fixes#780
Fixes#781
Fixes#782

Once again I could probably use some assistance with the appraisal gem but other than that should be OK.

Edit: This turned into a BIG refactor of the tests to split out support and I'll be honest my code isn't the tidiest. Still plenty of refactoring to go as I'm not sure I'm thrilled about how I've switched method definitions of Webpacker version numbers.

[AnatoliiD]

should be

Webpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configuration

[BookOfGreg]

Aaw whoops. Wondered why my test didn't pass XD Thanks

[BookOfGreg]

I've still got issues around Webpacker::Manifest.load, seems to be how I detect respond_to? may be wrong for that one.

[bradleesand]

The install generator won't work because entry_path was removed from webpacker config.

[bradleesand]

The new singleton implementation of Webpacker::Configuration no longer has entry_path. Instead, Webpacker::Configuration.source_path.join(Webpacker::Configuration.entry_path) becomes Webpacker.config.source_entry_path

[vikash-k-singh]

I have suggested these changes in issue #778 .Please have a look into it. I was trying to install react with rails and failed to run react installer command. I look into gem and found that method calling is not proper.

I fix this with minor changes in lib/generators/react/install_generator.rb file.

source_path is a instance method and was being called by class directly. Also entry_path was changed to source_entry_path in webpacker config. I have suggested these changes in #778

The main purpose for issue was #778 was to elevate exact issue and its solution along with it, as there was no fix until that time.

[basicBrogrammer]

Using webpacker 3.0.1 Webpacker.config.source_entry_path worked for me 👍

[vikash-k-singh]

It looks good, only a minor change suggested that webpack_configuration.entry_path should be changed to webpack_configuration.source_entry_path, as entry_path is renamed as source_entry_path in webpacker config file. File location for this file is lib/generators/react/install_generator.rb under method name javascript_dir.

[BookOfGreg]

I took on the code from #778 , thank you for that one! Makes it a little easier to update.

Currently working on trying to make the new config/webpacker.yml and it's env files play nicely with webpacker V1's paths.yml. It's putting my manifest in the webpacker.yml's output location with the paths.yml's manifest name so the webpacker helpers can't find the manifest.

Will take me some time to figure out all the subtleties of this before I'm done. Very welcome for someone to help with this!

Edit: Is there a nice way of changing the config files in test/dummy?

[vikash-k-singh]

In file lib/react/server_rendering/webpacker_manifest_container.rb, I am looking at a method name lookup_path at line no. 24, but I don't think it is defined anywhere in the both react-rails & webpacker gem. This will through an error once find_asset(logical_path) method is being called.

Please look into it! Are you guys loading any file with such variable?

Apart for this, hello @BookOfGreg, what kind of changes you wanna in test/dummy config files. Does it contains any Json data? Please share, hopefully I may change that!

[BookOfGreg]

Webpacker 1 and 2 have lookup_path, Webpacker 3 does not. and I'll have to compose a method of finding the actual path of a file.

[basicBrogrammer]

I'm not sure if this is related, but the component generator is putting my components in app/javascript/packs/components. To get this to work, I had to change the require context from 'components' to './components'

app/javascript/packs/application.js
var componentRequireContext = require.context('./components', true')

And also in my es6 components I have to require('../components/NameOfComponent')
Any thoughts?

[swrobel]

Seems like you might've messed up the indentation here

[bradleesand]

The <<- syntax allows the closing HEREDOC to be indented but it won't strip space from the content so this works just fine.

[willcosgrove]

I think this needs to be

defconfigWebpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configurationend

Because even in Webpacker 3, the Webpacker::Configuration constant is defined.

[tomwaddington]

Just ran into issue #778, and this fixed it for me. 👍

[BookOfGreg]

I'll admit that I'm at a bit of a dead end. On #779 I've got all of Travis passing except Rails 5 + Webpacker, but on that branch Webpack builds to public/packs but Webpacker looks in packs/packs instead. That is the case even if I peg Webpack at 1.15.x and Webpacker at ~> 1.0.
If a regular committer could help me out to get the tests passing on master then I could continue working here to get more Webpacker support.

We should really think about the maintainability of the Webpacker support as the manifest paths between versions seems really flakey and i'm not keen on the pattern of .responds_to? and defined? everywhere.

[BookOfGreg]

Master finally passes and this does mean Webpacker 3 support (And 1.1, 1.2, 2.0) on this branch.
I split out as much as I could do make it clearer but there is still a lot of mess I caused in the code, lots of refactoring to do.

[RiccardoMargiotta]

@BookOfGreg 😄

[rmosolgo]

Wow, this is really impressive! Thanks for all your work in digging in on this. It's nice to see that Webpacker has provided some APIs to support this kind of thing, too.

I see the green check, and the tests look good. So, what's still to go on this one?

[ArnonHongklay]

👍

[BookOfGreg]

This could go in to master it's current state as I believe it to be an improvement but frankly this could use some TLC and refactoring.

I'm certain all my MAJOR < 3 checks could be extracted into a class following the existing pattern.
There are probably better ways of integrating with the Webpacker.dev_server also.

I found some odd behaviour such as webpack 3 seems to read the wrong node_modules folder when loading in ../../../../react_ujs, so I used the actual npm package for it instead, that might cause some people in the wild issues that I can't even predict if they're messing with node context.

All these tweaks could be in future patches to this though.

Edit: Since I've modified the webpacker 1.1, 1.2 and 2 support, it would be good to see if it still works without bugs on an existing project for confirmation.

[Melonbwead]

You beast @BookOfGreg 👍 ❗️

[BookOfGreg]

I think we should probably merge this if we can be confident that it works. All existing tests pass. I can carry on with a tidy up beyond this.

Edit: apparently linting broke the tests. How odd. Will keep looking into it.

[BookOfGreg]

Tests fail on specific seeds:
bundle exec appraisal rails-5_no_sprockets_webpacker_3 rake test TESTOPTS="--seed=17880"

[BookOfGreg]

@rmosolgo If you like, and I'll keep PR'ing rubocops, test fixes and refactorings as patch-level changes after this. Be good to start getting some bug reports from folks so I know if this accidentally impacts anyone.

[rmosolgo]

That sounds great, let me go ahead with a release!

[gaiapunk]

I'm still having the #778 issue and was wondering if this latest merge had been release to fix it or if there is a workaround on my end that I can do before a fix is released, thanks for all the work you do!

[BookOfGreg]

@gaiapunk It's not been released as a gem yet but it's possible to use the master branch directly with:
gem 'react-rails', git: 'https://github.com/reactjs/react-rails.git', branch: 'master'

[BookOfGreg]

Fixes#776
Fixes#775
Fixes#778
Fixes#780
Fixes#781
Fixes#782

Once again I could probably use some assistance with the appraisal gem but other than that should be OK.

Edit: This turned into a BIG refactor of the tests to split out support and I'll be honest my code isn't the tidiest. Still plenty of refactoring to go as I'm not sure I'm thrilled about how I've switched method definitions of Webpacker version numbers.

[AnatoliiD]

should be

Webpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configuration

[BookOfGreg]

Aaw whoops. Wondered why my test didn't pass XD Thanks

[BookOfGreg]

I've still got issues around Webpacker::Manifest.load, seems to be how I detect respond_to? may be wrong for that one.

[bradleesand]

The install generator won't work because entry_path was removed from webpacker config.

[bradleesand]

The new singleton implementation of Webpacker::Configuration no longer has entry_path. Instead, Webpacker::Configuration.source_path.join(Webpacker::Configuration.entry_path) becomes Webpacker.config.source_entry_path

[vikash-k-singh]

I have suggested these changes in issue #778 .Please have a look into it. I was trying to install react with rails and failed to run react installer command. I look into gem and found that method calling is not proper.

I fix this with minor changes in lib/generators/react/install_generator.rb file.

source_path is a instance method and was being called by class directly. Also entry_path was changed to source_entry_path in webpacker config. I have suggested these changes in #778

The main purpose for issue was #778 was to elevate exact issue and its solution along with it, as there was no fix until that time.

[basicBrogrammer]

Using webpacker 3.0.1 Webpacker.config.source_entry_path worked for me 👍

[vikash-k-singh]

It looks good, only a minor change suggested that webpack_configuration.entry_path should be changed to webpack_configuration.source_entry_path, as entry_path is renamed as source_entry_path in webpacker config file. File location for this file is lib/generators/react/install_generator.rb under method name javascript_dir.

[BookOfGreg]

I took on the code from #778 , thank you for that one! Makes it a little easier to update.

Currently working on trying to make the new config/webpacker.yml and it's env files play nicely with webpacker V1's paths.yml. It's putting my manifest in the webpacker.yml's output location with the paths.yml's manifest name so the webpacker helpers can't find the manifest.

Will take me some time to figure out all the subtleties of this before I'm done. Very welcome for someone to help with this!

Edit: Is there a nice way of changing the config files in test/dummy?

[vikash-k-singh]

In file lib/react/server_rendering/webpacker_manifest_container.rb, I am looking at a method name lookup_path at line no. 24, but I don't think it is defined anywhere in the both react-rails & webpacker gem. This will through an error once find_asset(logical_path) method is being called.

Please look into it! Are you guys loading any file with such variable?

Apart for this, hello @BookOfGreg, what kind of changes you wanna in test/dummy config files. Does it contains any Json data? Please share, hopefully I may change that!

[BookOfGreg]

Webpacker 1 and 2 have lookup_path, Webpacker 3 does not. and I'll have to compose a method of finding the actual path of a file.

[basicBrogrammer]

I'm not sure if this is related, but the component generator is putting my components in app/javascript/packs/components. To get this to work, I had to change the require context from 'components' to './components'

app/javascript/packs/application.js
var componentRequireContext = require.context('./components', true')

And also in my es6 components I have to require('../components/NameOfComponent')
Any thoughts?

[swrobel]

Seems like you might've messed up the indentation here

[bradleesand]

The <<- syntax allows the closing HEREDOC to be indented but it won't strip space from the content so this works just fine.

[willcosgrove]

I think this needs to be

defconfigWebpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configurationend

Because even in Webpacker 3, the Webpacker::Configuration constant is defined.

[tomwaddington]

Just ran into issue #778, and this fixed it for me. 👍

[BookOfGreg]

I'll admit that I'm at a bit of a dead end. On #779 I've got all of Travis passing except Rails 5 + Webpacker, but on that branch Webpack builds to public/packs but Webpacker looks in packs/packs instead. That is the case even if I peg Webpack at 1.15.x and Webpacker at ~> 1.0.
If a regular committer could help me out to get the tests passing on master then I could continue working here to get more Webpacker support.

We should really think about the maintainability of the Webpacker support as the manifest paths between versions seems really flakey and i'm not keen on the pattern of .responds_to? and defined? everywhere.

[BookOfGreg]

Master finally passes and this does mean Webpacker 3 support (And 1.1, 1.2, 2.0) on this branch.
I split out as much as I could do make it clearer but there is still a lot of mess I caused in the code, lots of refactoring to do.

[RiccardoMargiotta]

@BookOfGreg 😄

[rmosolgo]

Wow, this is really impressive! Thanks for all your work in digging in on this. It's nice to see that Webpacker has provided some APIs to support this kind of thing, too.

I see the green check, and the tests look good. So, what's still to go on this one?

[ArnonHongklay]

👍

[BookOfGreg]

This could go in to master it's current state as I believe it to be an improvement but frankly this could use some TLC and refactoring.

I'm certain all my MAJOR < 3 checks could be extracted into a class following the existing pattern.
There are probably better ways of integrating with the Webpacker.dev_server also.

I found some odd behaviour such as webpack 3 seems to read the wrong node_modules folder when loading in ../../../../react_ujs, so I used the actual npm package for it instead, that might cause some people in the wild issues that I can't even predict if they're messing with node context.

All these tweaks could be in future patches to this though.

Edit: Since I've modified the webpacker 1.1, 1.2 and 2 support, it would be good to see if it still works without bugs on an existing project for confirmation.

[Melonbwead]

You beast @BookOfGreg 👍 ❗️

[BookOfGreg]

I think we should probably merge this if we can be confident that it works. All existing tests pass. I can carry on with a tidy up beyond this.

Edit: apparently linting broke the tests. How odd. Will keep looking into it.

[BookOfGreg]

Tests fail on specific seeds:
bundle exec appraisal rails-5_no_sprockets_webpacker_3 rake test TESTOPTS="--seed=17880"

[BookOfGreg]

@rmosolgo If you like, and I'll keep PR'ing rubocops, test fixes and refactorings as patch-level changes after this. Be good to start getting some bug reports from folks so I know if this accidentally impacts anyone.

[rmosolgo]

That sounds great, let me go ahead with a release!

[gaiapunk]

I'm still having the #778 issue and was wondering if this latest merge had been release to fix it or if there is a workaround on my end that I can do before a fix is released, thanks for all the work you do!

[BookOfGreg]

@gaiapunk It's not been released as a gem yet but it's possible to use the master branch directly with:
gem 'react-rails', git: 'https://github.com/reactjs/react-rails.git', branch: 'master'

[BookOfGreg]

Fixes#776
Fixes#775
Fixes#778
Fixes#780
Fixes#781
Fixes#782

Once again I could probably use some assistance with the appraisal gem but other than that should be OK.

Edit: This turned into a BIG refactor of the tests to split out support and I'll be honest my code isn't the tidiest. Still plenty of refactoring to go as I'm not sure I'm thrilled about how I've switched method definitions of Webpacker version numbers.

[AnatoliiD]

should be

Webpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configuration

[BookOfGreg]

Aaw whoops. Wondered why my test didn't pass XD Thanks

[BookOfGreg]

I've still got issues around Webpacker::Manifest.load, seems to be how I detect respond_to? may be wrong for that one.

[bradleesand]

The install generator won't work because entry_path was removed from webpacker config.

[bradleesand]

The new singleton implementation of Webpacker::Configuration no longer has entry_path. Instead, Webpacker::Configuration.source_path.join(Webpacker::Configuration.entry_path) becomes Webpacker.config.source_entry_path

[vikash-k-singh]

I have suggested these changes in issue #778 .Please have a look into it. I was trying to install react with rails and failed to run react installer command. I look into gem and found that method calling is not proper.

I fix this with minor changes in lib/generators/react/install_generator.rb file.

source_path is a instance method and was being called by class directly. Also entry_path was changed to source_entry_path in webpacker config. I have suggested these changes in #778

The main purpose for issue was #778 was to elevate exact issue and its solution along with it, as there was no fix until that time.

[basicBrogrammer]

Using webpacker 3.0.1 Webpacker.config.source_entry_path worked for me 👍

[vikash-k-singh]

It looks good, only a minor change suggested that webpack_configuration.entry_path should be changed to webpack_configuration.source_entry_path, as entry_path is renamed as source_entry_path in webpacker config file. File location for this file is lib/generators/react/install_generator.rb under method name javascript_dir.

[BookOfGreg]

I took on the code from #778 , thank you for that one! Makes it a little easier to update.

Currently working on trying to make the new config/webpacker.yml and it's env files play nicely with webpacker V1's paths.yml. It's putting my manifest in the webpacker.yml's output location with the paths.yml's manifest name so the webpacker helpers can't find the manifest.

Will take me some time to figure out all the subtleties of this before I'm done. Very welcome for someone to help with this!

Edit: Is there a nice way of changing the config files in test/dummy?

[vikash-k-singh]

In file lib/react/server_rendering/webpacker_manifest_container.rb, I am looking at a method name lookup_path at line no. 24, but I don't think it is defined anywhere in the both react-rails & webpacker gem. This will through an error once find_asset(logical_path) method is being called.

Please look into it! Are you guys loading any file with such variable?

Apart for this, hello @BookOfGreg, what kind of changes you wanna in test/dummy config files. Does it contains any Json data? Please share, hopefully I may change that!

[BookOfGreg]

Webpacker 1 and 2 have lookup_path, Webpacker 3 does not. and I'll have to compose a method of finding the actual path of a file.

[basicBrogrammer]

I'm not sure if this is related, but the component generator is putting my components in app/javascript/packs/components. To get this to work, I had to change the require context from 'components' to './components'

app/javascript/packs/application.js
var componentRequireContext = require.context('./components', true')

And also in my es6 components I have to require('../components/NameOfComponent')
Any thoughts?

[swrobel]

Seems like you might've messed up the indentation here

[bradleesand]

The <<- syntax allows the closing HEREDOC to be indented but it won't strip space from the content so this works just fine.

[willcosgrove]

I think this needs to be

defconfigWebpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configurationend

Because even in Webpacker 3, the Webpacker::Configuration constant is defined.

[tomwaddington]

Just ran into issue #778, and this fixed it for me. 👍

[BookOfGreg]

I'll admit that I'm at a bit of a dead end. On #779 I've got all of Travis passing except Rails 5 + Webpacker, but on that branch Webpack builds to public/packs but Webpacker looks in packs/packs instead. That is the case even if I peg Webpack at 1.15.x and Webpacker at ~> 1.0.
If a regular committer could help me out to get the tests passing on master then I could continue working here to get more Webpacker support.

We should really think about the maintainability of the Webpacker support as the manifest paths between versions seems really flakey and i'm not keen on the pattern of .responds_to? and defined? everywhere.

[BookOfGreg]

Master finally passes and this does mean Webpacker 3 support (And 1.1, 1.2, 2.0) on this branch.
I split out as much as I could do make it clearer but there is still a lot of mess I caused in the code, lots of refactoring to do.

[RiccardoMargiotta]

@BookOfGreg 😄

[rmosolgo]

Wow, this is really impressive! Thanks for all your work in digging in on this. It's nice to see that Webpacker has provided some APIs to support this kind of thing, too.

I see the green check, and the tests look good. So, what's still to go on this one?

[ArnonHongklay]

👍

[BookOfGreg]

This could go in to master it's current state as I believe it to be an improvement but frankly this could use some TLC and refactoring.

I'm certain all my MAJOR < 3 checks could be extracted into a class following the existing pattern.
There are probably better ways of integrating with the Webpacker.dev_server also.

I found some odd behaviour such as webpack 3 seems to read the wrong node_modules folder when loading in ../../../../react_ujs, so I used the actual npm package for it instead, that might cause some people in the wild issues that I can't even predict if they're messing with node context.

All these tweaks could be in future patches to this though.

Edit: Since I've modified the webpacker 1.1, 1.2 and 2 support, it would be good to see if it still works without bugs on an existing project for confirmation.

[Melonbwead]

You beast @BookOfGreg 👍 ❗️

[BookOfGreg]

I think we should probably merge this if we can be confident that it works. All existing tests pass. I can carry on with a tidy up beyond this.

Edit: apparently linting broke the tests. How odd. Will keep looking into it.

[BookOfGreg]

Tests fail on specific seeds:
bundle exec appraisal rails-5_no_sprockets_webpacker_3 rake test TESTOPTS="--seed=17880"

[BookOfGreg]

@rmosolgo If you like, and I'll keep PR'ing rubocops, test fixes and refactorings as patch-level changes after this. Be good to start getting some bug reports from folks so I know if this accidentally impacts anyone.

[rmosolgo]

That sounds great, let me go ahead with a release!

[gaiapunk]

I'm still having the #778 issue and was wondering if this latest merge had been release to fix it or if there is a workaround on my end that I can do before a fix is released, thanks for all the work you do!

[BookOfGreg]

@gaiapunk It's not been released as a gem yet but it's possible to use the master branch directly with:
gem 'react-rails', git: 'https://github.com/reactjs/react-rails.git', branch: 'master'

[BookOfGreg]

Fixes#776
Fixes#775
Fixes#778
Fixes#780
Fixes#781
Fixes#782

Once again I could probably use some assistance with the appraisal gem but other than that should be OK.

Edit: This turned into a BIG refactor of the tests to split out support and I'll be honest my code isn't the tidiest. Still plenty of refactoring to go as I'm not sure I'm thrilled about how I've switched method definitions of Webpacker version numbers.

[AnatoliiD]

should be

Webpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configuration

[BookOfGreg]

Aaw whoops. Wondered why my test didn't pass XD Thanks

[BookOfGreg]

I've still got issues around Webpacker::Manifest.load, seems to be how I detect respond_to? may be wrong for that one.

[bradleesand]

The install generator won't work because entry_path was removed from webpacker config.

[bradleesand]

The new singleton implementation of Webpacker::Configuration no longer has entry_path. Instead, Webpacker::Configuration.source_path.join(Webpacker::Configuration.entry_path) becomes Webpacker.config.source_entry_path

[vikash-k-singh]

I have suggested these changes in issue #778 .Please have a look into it. I was trying to install react with rails and failed to run react installer command. I look into gem and found that method calling is not proper.

I fix this with minor changes in lib/generators/react/install_generator.rb file.

source_path is a instance method and was being called by class directly. Also entry_path was changed to source_entry_path in webpacker config. I have suggested these changes in #778

The main purpose for issue was #778 was to elevate exact issue and its solution along with it, as there was no fix until that time.

[basicBrogrammer]

Using webpacker 3.0.1 Webpacker.config.source_entry_path worked for me 👍

[vikash-k-singh]

It looks good, only a minor change suggested that webpack_configuration.entry_path should be changed to webpack_configuration.source_entry_path, as entry_path is renamed as source_entry_path in webpacker config file. File location for this file is lib/generators/react/install_generator.rb under method name javascript_dir.

[BookOfGreg]

I took on the code from #778 , thank you for that one! Makes it a little easier to update.

Currently working on trying to make the new config/webpacker.yml and it's env files play nicely with webpacker V1's paths.yml. It's putting my manifest in the webpacker.yml's output location with the paths.yml's manifest name so the webpacker helpers can't find the manifest.

Will take me some time to figure out all the subtleties of this before I'm done. Very welcome for someone to help with this!

Edit: Is there a nice way of changing the config files in test/dummy?

[vikash-k-singh]

In file lib/react/server_rendering/webpacker_manifest_container.rb, I am looking at a method name lookup_path at line no. 24, but I don't think it is defined anywhere in the both react-rails & webpacker gem. This will through an error once find_asset(logical_path) method is being called.

Please look into it! Are you guys loading any file with such variable?

Apart for this, hello @BookOfGreg, what kind of changes you wanna in test/dummy config files. Does it contains any Json data? Please share, hopefully I may change that!

[BookOfGreg]

Webpacker 1 and 2 have lookup_path, Webpacker 3 does not. and I'll have to compose a method of finding the actual path of a file.

[basicBrogrammer]

I'm not sure if this is related, but the component generator is putting my components in app/javascript/packs/components. To get this to work, I had to change the require context from 'components' to './components'

app/javascript/packs/application.js
var componentRequireContext = require.context('./components', true')

And also in my es6 components I have to require('../components/NameOfComponent')
Any thoughts?

[swrobel]

Seems like you might've messed up the indentation here

[bradleesand]

The <<- syntax allows the closing HEREDOC to be indented but it won't strip space from the content so this works just fine.

[willcosgrove]

I think this needs to be

defconfigWebpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configurationend

Because even in Webpacker 3, the Webpacker::Configuration constant is defined.

[tomwaddington]

Just ran into issue #778, and this fixed it for me. 👍

[BookOfGreg]

I'll admit that I'm at a bit of a dead end. On #779 I've got all of Travis passing except Rails 5 + Webpacker, but on that branch Webpack builds to public/packs but Webpacker looks in packs/packs instead. That is the case even if I peg Webpack at 1.15.x and Webpacker at ~> 1.0.
If a regular committer could help me out to get the tests passing on master then I could continue working here to get more Webpacker support.

We should really think about the maintainability of the Webpacker support as the manifest paths between versions seems really flakey and i'm not keen on the pattern of .responds_to? and defined? everywhere.

[BookOfGreg]

Master finally passes and this does mean Webpacker 3 support (And 1.1, 1.2, 2.0) on this branch.
I split out as much as I could do make it clearer but there is still a lot of mess I caused in the code, lots of refactoring to do.

[RiccardoMargiotta]

@BookOfGreg 😄

[rmosolgo]

Wow, this is really impressive! Thanks for all your work in digging in on this. It's nice to see that Webpacker has provided some APIs to support this kind of thing, too.

I see the green check, and the tests look good. So, what's still to go on this one?

[ArnonHongklay]

👍

[BookOfGreg]

This could go in to master it's current state as I believe it to be an improvement but frankly this could use some TLC and refactoring.

I'm certain all my MAJOR < 3 checks could be extracted into a class following the existing pattern.
There are probably better ways of integrating with the Webpacker.dev_server also.

I found some odd behaviour such as webpack 3 seems to read the wrong node_modules folder when loading in ../../../../react_ujs, so I used the actual npm package for it instead, that might cause some people in the wild issues that I can't even predict if they're messing with node context.

All these tweaks could be in future patches to this though.

Edit: Since I've modified the webpacker 1.1, 1.2 and 2 support, it would be good to see if it still works without bugs on an existing project for confirmation.

[Melonbwead]

You beast @BookOfGreg 👍 ❗️

[BookOfGreg]

I think we should probably merge this if we can be confident that it works. All existing tests pass. I can carry on with a tidy up beyond this.

Edit: apparently linting broke the tests. How odd. Will keep looking into it.

[BookOfGreg]

Tests fail on specific seeds:
bundle exec appraisal rails-5_no_sprockets_webpacker_3 rake test TESTOPTS="--seed=17880"

[BookOfGreg]

@rmosolgo If you like, and I'll keep PR'ing rubocops, test fixes and refactorings as patch-level changes after this. Be good to start getting some bug reports from folks so I know if this accidentally impacts anyone.

[rmosolgo]

That sounds great, let me go ahead with a release!

[gaiapunk]

I'm still having the #778 issue and was wondering if this latest merge had been release to fix it or if there is a workaround on my end that I can do before a fix is released, thanks for all the work you do!

[BookOfGreg]

@gaiapunk It's not been released as a gem yet but it's possible to use the master branch directly with:
gem 'react-rails', git: 'https://github.com/reactjs/react-rails.git', branch: 'master'

[BookOfGreg]

Fixes#776
Fixes#775
Fixes#778
Fixes#780
Fixes#781
Fixes#782

Once again I could probably use some assistance with the appraisal gem but other than that should be OK.

Edit: This turned into a BIG refactor of the tests to split out support and I'll be honest my code isn't the tidiest. Still plenty of refactoring to go as I'm not sure I'm thrilled about how I've switched method definitions of Webpacker version numbers.

[AnatoliiD]

should be

Webpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configuration

[BookOfGreg]

Aaw whoops. Wondered why my test didn't pass XD Thanks

[BookOfGreg]

I've still got issues around Webpacker::Manifest.load, seems to be how I detect respond_to? may be wrong for that one.

[bradleesand]

The install generator won't work because entry_path was removed from webpacker config.

[bradleesand]

The new singleton implementation of Webpacker::Configuration no longer has entry_path. Instead, Webpacker::Configuration.source_path.join(Webpacker::Configuration.entry_path) becomes Webpacker.config.source_entry_path

[vikash-k-singh]

I have suggested these changes in issue #778 .Please have a look into it. I was trying to install react with rails and failed to run react installer command. I look into gem and found that method calling is not proper.

I fix this with minor changes in lib/generators/react/install_generator.rb file.

source_path is a instance method and was being called by class directly. Also entry_path was changed to source_entry_path in webpacker config. I have suggested these changes in #778

The main purpose for issue was #778 was to elevate exact issue and its solution along with it, as there was no fix until that time.

[basicBrogrammer]

Using webpacker 3.0.1 Webpacker.config.source_entry_path worked for me 👍

[vikash-k-singh]

It looks good, only a minor change suggested that webpack_configuration.entry_path should be changed to webpack_configuration.source_entry_path, as entry_path is renamed as source_entry_path in webpacker config file. File location for this file is lib/generators/react/install_generator.rb under method name javascript_dir.

[BookOfGreg]

I took on the code from #778 , thank you for that one! Makes it a little easier to update.

Currently working on trying to make the new config/webpacker.yml and it's env files play nicely with webpacker V1's paths.yml. It's putting my manifest in the webpacker.yml's output location with the paths.yml's manifest name so the webpacker helpers can't find the manifest.

Will take me some time to figure out all the subtleties of this before I'm done. Very welcome for someone to help with this!

Edit: Is there a nice way of changing the config files in test/dummy?

[vikash-k-singh]

In file lib/react/server_rendering/webpacker_manifest_container.rb, I am looking at a method name lookup_path at line no. 24, but I don't think it is defined anywhere in the both react-rails & webpacker gem. This will through an error once find_asset(logical_path) method is being called.

Please look into it! Are you guys loading any file with such variable?

Apart for this, hello @BookOfGreg, what kind of changes you wanna in test/dummy config files. Does it contains any Json data? Please share, hopefully I may change that!

[BookOfGreg]

Webpacker 1 and 2 have lookup_path, Webpacker 3 does not. and I'll have to compose a method of finding the actual path of a file.

[basicBrogrammer]

I'm not sure if this is related, but the component generator is putting my components in app/javascript/packs/components. To get this to work, I had to change the require context from 'components' to './components'

app/javascript/packs/application.js
var componentRequireContext = require.context('./components', true')

And also in my es6 components I have to require('../components/NameOfComponent')
Any thoughts?

[swrobel]

Seems like you might've messed up the indentation here

[bradleesand]

The <<- syntax allows the closing HEREDOC to be indented but it won't strip space from the content so this works just fine.

[willcosgrove]

I think this needs to be

defconfigWebpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configurationend

Because even in Webpacker 3, the Webpacker::Configuration constant is defined.

[tomwaddington]

Just ran into issue #778, and this fixed it for me. 👍

[BookOfGreg]

I'll admit that I'm at a bit of a dead end. On #779 I've got all of Travis passing except Rails 5 + Webpacker, but on that branch Webpack builds to public/packs but Webpacker looks in packs/packs instead. That is the case even if I peg Webpack at 1.15.x and Webpacker at ~> 1.0.
If a regular committer could help me out to get the tests passing on master then I could continue working here to get more Webpacker support.

We should really think about the maintainability of the Webpacker support as the manifest paths between versions seems really flakey and i'm not keen on the pattern of .responds_to? and defined? everywhere.

[BookOfGreg]

Master finally passes and this does mean Webpacker 3 support (And 1.1, 1.2, 2.0) on this branch.
I split out as much as I could do make it clearer but there is still a lot of mess I caused in the code, lots of refactoring to do.

[RiccardoMargiotta]

@BookOfGreg 😄

[rmosolgo]

Wow, this is really impressive! Thanks for all your work in digging in on this. It's nice to see that Webpacker has provided some APIs to support this kind of thing, too.

I see the green check, and the tests look good. So, what's still to go on this one?

[ArnonHongklay]

👍

[BookOfGreg]

This could go in to master it's current state as I believe it to be an improvement but frankly this could use some TLC and refactoring.

I'm certain all my MAJOR < 3 checks could be extracted into a class following the existing pattern.
There are probably better ways of integrating with the Webpacker.dev_server also.

I found some odd behaviour such as webpack 3 seems to read the wrong node_modules folder when loading in ../../../../react_ujs, so I used the actual npm package for it instead, that might cause some people in the wild issues that I can't even predict if they're messing with node context.

All these tweaks could be in future patches to this though.

Edit: Since I've modified the webpacker 1.1, 1.2 and 2 support, it would be good to see if it still works without bugs on an existing project for confirmation.

[Melonbwead]

You beast @BookOfGreg 👍 ❗️

[BookOfGreg]

I think we should probably merge this if we can be confident that it works. All existing tests pass. I can carry on with a tidy up beyond this.

Edit: apparently linting broke the tests. How odd. Will keep looking into it.

[BookOfGreg]

Tests fail on specific seeds:
bundle exec appraisal rails-5_no_sprockets_webpacker_3 rake test TESTOPTS="--seed=17880"

[BookOfGreg]

@rmosolgo If you like, and I'll keep PR'ing rubocops, test fixes and refactorings as patch-level changes after this. Be good to start getting some bug reports from folks so I know if this accidentally impacts anyone.

[rmosolgo]

That sounds great, let me go ahead with a release!

[gaiapunk]

I'm still having the #778 issue and was wondering if this latest merge had been release to fix it or if there is a workaround on my end that I can do before a fix is released, thanks for all the work you do!

[BookOfGreg]

@gaiapunk It's not been released as a gem yet but it's possible to use the master branch directly with:
gem 'react-rails', git: 'https://github.com/reactjs/react-rails.git', branch: 'master'

[BookOfGreg]

Fixes#776
Fixes#775
Fixes#778
Fixes#780
Fixes#781
Fixes#782

Once again I could probably use some assistance with the appraisal gem but other than that should be OK.

Edit: This turned into a BIG refactor of the tests to split out support and I'll be honest my code isn't the tidiest. Still plenty of refactoring to go as I'm not sure I'm thrilled about how I've switched method definitions of Webpacker version numbers.

[AnatoliiD]

should be

Webpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configuration

[BookOfGreg]

Aaw whoops. Wondered why my test didn't pass XD Thanks

[BookOfGreg]

I've still got issues around Webpacker::Manifest.load, seems to be how I detect respond_to? may be wrong for that one.

[bradleesand]

The install generator won't work because entry_path was removed from webpacker config.

[bradleesand]

The new singleton implementation of Webpacker::Configuration no longer has entry_path. Instead, Webpacker::Configuration.source_path.join(Webpacker::Configuration.entry_path) becomes Webpacker.config.source_entry_path

[vikash-k-singh]

I have suggested these changes in issue #778 .Please have a look into it. I was trying to install react with rails and failed to run react installer command. I look into gem and found that method calling is not proper.

I fix this with minor changes in lib/generators/react/install_generator.rb file.

source_path is a instance method and was being called by class directly. Also entry_path was changed to source_entry_path in webpacker config. I have suggested these changes in #778

The main purpose for issue was #778 was to elevate exact issue and its solution along with it, as there was no fix until that time.

[basicBrogrammer]

Using webpacker 3.0.1 Webpacker.config.source_entry_path worked for me 👍

[vikash-k-singh]

It looks good, only a minor change suggested that webpack_configuration.entry_path should be changed to webpack_configuration.source_entry_path, as entry_path is renamed as source_entry_path in webpacker config file. File location for this file is lib/generators/react/install_generator.rb under method name javascript_dir.

[BookOfGreg]

I took on the code from #778 , thank you for that one! Makes it a little easier to update.

Currently working on trying to make the new config/webpacker.yml and it's env files play nicely with webpacker V1's paths.yml. It's putting my manifest in the webpacker.yml's output location with the paths.yml's manifest name so the webpacker helpers can't find the manifest.

Will take me some time to figure out all the subtleties of this before I'm done. Very welcome for someone to help with this!

Edit: Is there a nice way of changing the config files in test/dummy?

[vikash-k-singh]

In file lib/react/server_rendering/webpacker_manifest_container.rb, I am looking at a method name lookup_path at line no. 24, but I don't think it is defined anywhere in the both react-rails & webpacker gem. This will through an error once find_asset(logical_path) method is being called.

Please look into it! Are you guys loading any file with such variable?

Apart for this, hello @BookOfGreg, what kind of changes you wanna in test/dummy config files. Does it contains any Json data? Please share, hopefully I may change that!

[BookOfGreg]

Webpacker 1 and 2 have lookup_path, Webpacker 3 does not. and I'll have to compose a method of finding the actual path of a file.

[basicBrogrammer]

I'm not sure if this is related, but the component generator is putting my components in app/javascript/packs/components. To get this to work, I had to change the require context from 'components' to './components'

app/javascript/packs/application.js
var componentRequireContext = require.context('./components', true')

And also in my es6 components I have to require('../components/NameOfComponent')
Any thoughts?

[swrobel]

Seems like you might've messed up the indentation here

[bradleesand]

The <<- syntax allows the closing HEREDOC to be indented but it won't strip space from the content so this works just fine.

[willcosgrove]

I think this needs to be

defconfigWebpacker.respond_to?(:config) ? Webpacker.config : Webpacker::Configurationend

Because even in Webpacker 3, the Webpacker::Configuration constant is defined.

[tomwaddington]

Just ran into issue #778, and this fixed it for me. 👍

[BookOfGreg]

I'll admit that I'm at a bit of a dead end. On #779 I've got all of Travis passing except Rails 5 + Webpacker, but on that branch Webpack builds to public/packs but Webpacker looks in packs/packs instead. That is the case even if I peg Webpack at 1.15.x and Webpacker at ~> 1.0.
If a regular committer could help me out to get the tests passing on master then I could continue working here to get more Webpacker support.

We should really think about the maintainability of the Webpacker support as the manifest paths between versions seems really flakey and i'm not keen on the pattern of .responds_to? and defined? everywhere.

[BookOfGreg]

Master finally passes and this does mean Webpacker 3 support (And 1.1, 1.2, 2.0) on this branch.
I split out as much as I could do make it clearer but there is still a lot of mess I caused in the code, lots of refactoring to do.

[RiccardoMargiotta]

@BookOfGreg 😄

[rmosolgo]

Wow, this is really impressive! Thanks for all your work in digging in on this. It's nice to see that Webpacker has provided some APIs to support this kind of thing, too.

I see the green check, and the tests look good. So, what's still to go on this one?

[ArnonHongklay]

👍

[BookOfGreg]

This could go in to master it's current state as I believe it to be an improvement but frankly this could use some TLC and refactoring.

I'm certain all my MAJOR < 3 checks could be extracted into a class following the existing pattern.
There are probably better ways of integrating with the Webpacker.dev_server also.

I found some odd behaviour such as webpack 3 seems to read the wrong node_modules folder when loading in ../../../../react_ujs, so I used the actual npm package for it instead, that might cause some people in the wild issues that I can't even predict if they're messing with node context.

All these tweaks could be in future patches to this though.

Edit: Since I've modified the webpacker 1.1, 1.2 and 2 support, it would be good to see if it still works without bugs on an existing project for confirmation.

[Melonbwead]

You beast @BookOfGreg 👍 ❗️

[BookOfGreg]

I think we should probably merge this if we can be confident that it works. All existing tests pass. I can carry on with a tidy up beyond this.

Edit: apparently linting broke the tests. How odd. Will keep looking into it.

[BookOfGreg]

Tests fail on specific seeds:
bundle exec appraisal rails-5_no_sprockets_webpacker_3 rake test TESTOPTS="--seed=17880"

[BookOfGreg]

@rmosolgo If you like, and I'll keep PR'ing rubocops, test fixes and refactorings as patch-level changes after this. Be good to start getting some bug reports from folks so I know if this accidentally impacts anyone.

[rmosolgo]

That sounds great, let me go ahead with a release!

[gaiapunk]

I'm still having the #778 issue and was wondering if this latest merge had been release to fix it or if there is a workaround on my end that I can do before a fix is released, thanks for all the work you do!

[BookOfGreg]

@gaiapunk It's not been released as a gem yet but it's possible to use the master branch directly with:
gem 'react-rails', git: 'https://github.com/reactjs/react-rails.git', branch: 'master'