Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Refactor code to allow splitting bundle data extraction from UI code #40

Closed
wants to merge 27 commits into from

Conversation

valscion
Copy link
Member

The goal of this PR is to split the UI code into a separate package that can be leveraged by the current webpack-bundle-analyzer code.

See #32 for more reasoning

That file is the only thing that references the moved function
It was confusing to have both src/bin/analyzer.js and src/analyzer.js
when the latter only provided one function used elsewhere.
This helps reduce the mental overhead of all files residing under the
same directory level in src/
They are only used from viewer/getViewerData so grouping them together
helps in finding out what files reference each other
@valscion
Copy link
Member Author

So far I have moved files around src/ directory so that I could get a better focus on what files require each other. I hope that these file locations were internal to this module and weren't part of the public API.

Each commit up to this point has passed tests, which can be seen from https://travis-ci.org/valscion/webpack-bundle-analyzer/builds, too

I was wondering, why do some files require other modules via the lib/ build directory? I understand why the tests might want to do that, but not why the source files do that.

I'll change all the non-test requires to point to files inside src directory. It should be easy-ish to revert, if there is a good reason to still require from lib.

I don't really understand why these modules were imported from the built
`lib/` directory before
This is a partial step in refactoring the whole, as it allows me to
focus on less things at once.

I don't know yet the complete file structure of what the source will
look like in the end, but this is a useful, safe refactoring step on the
way to get there 😄
Now all the call sites should provide a logger instance to the distinct
functions
It makes it a bit difficult to follow what module is in charge of
providing defaults if every one of them is overly protective and knows
how instantiate their loggers.

Defining clearer boundaries for the individual pieces allows for safer
refactoring, and should something fail, the parameter invariant should
be triggered to get a good bug report.
@valscion
Copy link
Member Author

I've first had to refactor the current code to get it to a state where I feel more confident in splitting out distinct parts. I try to keep the steps small so that following along one commit a time should make for a pleasant review experience.

Please let me know if you have trouble understanding anything I've done. I'm so far targeting to create breathing room for the separation, and I'm not yet 100% sure what the code will look like once I'm done.

I'm not entirely sure we even need to expose the `start` function to the
outside world at all.

This commit seems to be the origin of the deprected comment:
d644450

It might be safe to remove this export altogether, but I'd rather err on
the safe side to avoid breaking dependents code
It should not be the responsibility of the startServer function to also
know how to load the final data. It makes for cleaner code to plumb it
in from the outside so that the server related code can focus on just
launching the server and nothing more.
Now that the parameter clearly is there only to provide a configuration
hook for server related things, it's quite fine to allow the
serverOptions parameter to be missing as sane defaults can be set
@valscion
Copy link
Member Author

valscion commented Jan 21, 2017

EDIT:

Let's ignore this comment for now, we don't need to address it in this PR


At this point in the refactoring, I'm starting to miss the lovely flow types as all the various option parameter defaults and plumbing is starting to get confusing again.

What is your stance on static typing, @th0r? Would you be fine if I'd introduce some gradual typing to the webpack-bundle-analyzer codebase so that touching some function parameters wouldn't so easily break accidentally?

This file structure is starting to make more sense, bit by bit...
I might've gone a bit overboard with commits
0369975 and
2e92a49

This approach looks better, as now startServer looks a bit like
generateReport does.
@valscion
Copy link
Member Author

valscion commented Jan 21, 2017

Huh? Why did the 640098b commit build pass? https://travis-ci.org/th0r/webpack-bundle-analyzer/builds/194015131

It should've been broken! It referenced a file that got moved to a different folder:

// test/viewer/parseUtils.js
const { parseBundle } = require('../../lib/viewer/parseUtils');

😕 It should've been this:

const { parseBundle } = require('../../lib/chartData/parseUtils');

Really strange.

EDIT: Oh, mocha didn't pick up those tests at all as they were inside a subdirectory.... oh well. d86ea96

Seems like mocha didn't pick up these tests at all when they were moved
to a subfolder 😱
This way it is more clear what is the importable file for the outside
modules. Other modules should not import `getViewerData`, `parseUtils`
or `tree` files, as they should be considered internal to the
getChartData function.
This makes it more clear that the entire file exports only one function
This is more clear as to what is passed down to `getViewerData` from
`getChartData` function than relying on argument order and splatting
them.
@valscion
Copy link
Member Author

valscion commented Jan 21, 2017

What's the reason behind not requiring bundleDir to be always present, but having this check in src/viewer.js?

      if (!path.isAbsolute(reportFilepath)) {
        reportFilepath = path.resolve(bundleDir || process.cwd(), reportFilepath);
      }

I see you also check for its presence in src/analyzer.js module's getViewerData function:

  // Trying to parse bundle assets and get real module sizes if `bundleDir` is provided
  let parsedModuleSizes = null;
  let bundlesSources = {};
  let parsedModules = {};

  if (bundleDir) {

The process.cwd() defaulting seems to have been first introduced in a commit adding "static report" mode and the protective if(bundleDir) in getViewerData has been there ever since the initial commit.

The bundleDir value can come in via only two ways, that are more easily spotted in this PR:

  1. From the plugin instance getting it from this.compiler.outputPath
  2. From the CLI arguments or inferring it from the bundle stat file location

It seems like there shouldn't be any scenario where the bundleDir should be null or undefined after these steps. If I haven't misunderstood, the this.compiler.outputPath comes from output.path webpack config value which in turn is required by webpack to work.

So it should be fairly safe to remove the overly protective check from the code. It would make my life a bit easier as I'd have one less thing to worry about 😄

The `bundleDir` value can come in via only two ways:

1. From the **plugin instance** getting it from
   `this.compiler.outputPath`

2. From the **CLI** arguments or inferring it from the bundle stat file
   location

It seems like there shouldn't be any scenario where the `bundleDir`
should be `null` or `undefined` after these steps. If I haven't
misunderstood, the `this.compiler.outputPath` comes from `output.path`
webpack config value which in turn is required by webpack to be set.

So it should be fairly safe to remove the overly protective check from
the code.
@valscion
Copy link
Member Author

I went and removed the bundleDir optionality in the last commit. As the indentation changed, it might be best viewed with whitespace turned off using the special ?w=1 query parameter in the URL: 21385bf?w=1

Now we're talking! Now the data mangling is happening at the very edges
of the code, so separating the UI related code below will finally be
possible! 💃
@valscion valscion changed the title Split bundle data extraction from UI code Refactor code to allow splitting bundle data extraction from UI code Jan 21, 2017
@valscion
Copy link
Member Author

I'll tackle the actual splitting of UI and chart data extraction in a new PR so that we can focus on only the refactoring here. Otherwise this PR risks becoming too big and the review would get waaaaay too cumbersome.

Let me know if you have any questions! I'm very open in discussing any approaches I've taken. If you've got critique, I want to hear it! 😄

@valscion
Copy link
Member Author

Hiya @th0r, have you had any time to look into this? If you're having trouble maintaining this project, I could help you. It'd be a shame for this project to wither as it is so useful to many :)

@valscion
Copy link
Member Author

Hi @th0r, would you be able to review this? It'd be nice to get this refactor merged so that there would be less merge conflicts in the future.

I'm fairly confident that this does not introduce any changes or backwards-incompatibilities. webpack-bundle-analyzer has never seemed to support reaching to the insides of this plugin, so moving things around should also be safe.

I would like to continue figuring out a good approach on splitting the view code altogether, but that I would do in another place :)

The current file hierarchy looks like this now:

webpack-bundle-analyzer/
├── client
│   ├── components
│   └── vendor
├── lib
│   ├── bin
│   ├── chartData
│   └── viewer
├── node_modules
├── public
├── src
│   ├── bin
│   ├── chartData
│   └── viewer
├── test
│   ├── bundles
│   ├── src
│   └── stats
└── views

The only things reaching into top-level client, public and views folders are the root-level webpack.config.js and code under src/viewer.

Those folders and the webpack config form the implementation of the current treemap reporter.

With this in place I have a much better situation with figuring out how to split the reporter away from the data generation.

@valscion
Copy link
Member Author

I'll close this in favor of #81 that features a much smaller amount of code changes.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants