Remove Rx from core

[jhusain]

We plan to evaluate removing RX from the core of Falcor. There are several things that motivate this decision:

1. speed
2. debuggability
3. file size

We will be doing some performance analysis over the next few weeks.

@yunong

[yunong]

Sounds good. I can help with performance analysis as needed.

[trxcllnt]

@jhusain @ktrott how'd we decide on this? should I stop work on integrating RxJS-Next?

[ewinslow]

+1 on filesize concerns. My eyes went like O_O when I saw how large Falcor is right now.

[ktrott]

@trxcllnt No decision has been made yet. You should proceed with your work as we're using that to evaluate the performance of the ReactiveX/RxJS Next project so that work is tied to both projects.

[trxcllnt]

The preliminary upgrade to RxJS-Next is done and available in my forks:

https://github.com/trxcllnt/falcor/tree/rx-next
https://github.com/trxcllnt/falcor-router/tree/rx-next

You’ll have to manually build RxJS-Next, since we’re not presently committing the “dist” folder.

For falcor:

~ cd ./node_modules/RxJS-Next
~ npm i

For falcor-router (have to build both falcor and falcor-router’s RxJS-Next packages):

~ cd ./node_modules/RxJS-Next
~ npm i
~ cd ../falcor
~ npm i
~ cd node_modules/RxJS-Next
~ npm i

[ThePrimeagen]

I will be finishing up the get request cycle to have rx completely removed from it. This will be second step towards NoRX in falcor. #506

[tivac]

I was curious about size savings, so did a quick test using browserify.

-norx files had rx/dist/rx and rx/dist/rx.aggregates excluded from the browserify build.
.min files have been minified with uglify-js
.gz files were gzipped with gzipme.

513,045 falcor.js
278,833 falcor.min.js
276,966 falcor-norx.js
161,130 falcor-norx.min.js
 98,226 falcor.js.gz
 57,624 falcor-norx.js.gz
 54,552 falcor.min.js.gz
 32,932 falcor-norx.min.js.gz

Obviously this doesn't take into account the size of whatever would replace rx.js, but it is interesting IMO.

[ThePrimeagen]

@tivac norx does not remove rx, just rx from the get requesting / batching / across flight deduping. Rx itself is about 100+k when minified. The falcor library should be about 160k minified (i am surprised at the 278k minified, there must be double Rx inclusion).

Our goal would be to remove both path syntax / rx and have our library down to about ~50k, minified. I think this goal is achievable soon.

[tivac]

@michaelbpaulson That's not from your norx branch. That's me excluding every rx-related module in my browserify config.

The source file is just require("falcor");, then I point browserify at it and see what happens :)

[ThePrimeagen]

The last of the required changes will be in this issue: #604
Once that is complete then we can remove Rx completely.

[ThePrimeagen]

Now Rx itself can be removed from the dependencies. The removal will take a moment due to unit test upgrade.

[tivac]

Did some new filesize comparisons of falcor@master vs falcor@0.1.14 now that #606 has landed.

• .min files are compressed with latest Uglify2
• .jsgz are compressed using gzip-size-cli
• Sizes calculated by running browserify against a file that simply contains require("falcor");

File	Size
falcor-0.1.14.js	499.41 kB
falcor-0.1.14.min.js	272.37 kB
falcor-master.js	250.17 kB
falcor-master.min.js	141.37 kB
falcor-0.1.14.jsgz	93.91 kB
falcor-0.1.14.min.jsgz	52.75 kB
falcor-master.jsgz	51.51 kB
falcor-master.min.jsgz	28.09 kB

About half the size, kudos @michaelbpaulson! 👍

[ThePrimeagen]

@tivac I don't think you are testing the correct file. A simple gulp dist shows falcor.browser.min.js being 68k on master right now. falcor.all.min.js is not the one to test as it includes falcor-router.

[tivac]

I mentioned my method in the post:

Sizes calculated by running browserify against a file that simply contains require("falcor");

Since we use falcor via browserify that test is at least accurate for our needs!