Pure render for "variable" components

[jtrein]

This is the base part of work to get OpenLawForm on large agreements to be performant. In my testing this does alleviate some performance issues with the large agreements (e.g. LSTA)! There is more work to be done in our app, specifically, but that's not really the business of this PR.

Work done

• Moves the calculation of Openlaw-specific props up a level to InputRenderer
• executionResult and variable are removed as prop dependencies, because they cannot be compared in order to not render, because they are always new objects.
• Any props that can be cached, or made to be easily compared, is completed.
• All variable components are now React.PureComponents

@jdville03 Can you give this a test-run in the openlaw web app, or the example app, to make sure I haven't damaged anything? In addition to my own testing, I've run npm test and sbt test (the sbt tests that are failing for sbt aren't b/c of this) on the openlaw app side with these components.

Note: If you're testing in openlaw web app, after using yarn link, you'll probably have to uncomment flow before building, otherwise there will be some incorrect flow errors.

[jdville03]

Sure I'll test this out today.

[jdville03]

So far I've found regression with the Identity Variable (testing in openlaw web app). The field clears after typing the @ in an email. I left the devtools open in the video below in case helpful to see the component props and state.

https://www.useloom.com/share/23ac85d96b81411cab39b6333134e7d1

[jdville03]

There is also regression with the YesNo variables.

This first set of videos shows the expected behavior and regression when they're used in a deal. Expected behavior (the conditional question remains in the deal agreement form when it's clicked on): https://www.useloom.com/share/9d8531885f384e6ab2f71e8a6235223b Regression (the conditional question gets removed from the deal agreement form when it's clicked on): https://www.useloom.com/share/03c071adfed2450e9c25894a69ab8ad8

This second set of videos shows the difference in the context of a flow form. Expected behavior (the radio is rendered correctly as checked when clicking on it, leaving the form page, and coming back to that form page): https://www.useloom.com/share/059b9e1f0086471198e7f42672286dce Regression (the radio button is no longer checked when clicking on it, leaving the form page, and coming back to that form page): https://www.useloom.com/share/f3186f112f7644159513d75e6d478d08

[jdville03]

These tests are new failures for me in the openlaw web app when I'm running with the test local package in this pr:

org.adridadou.openlaw.demo.EmployeeOnboardingE2ESpec
org.adridadou.openlaw.e2e.StopContractE2ESpec
org.adridadou.openlaw.e2e.EthereumCallE2ESpec

They all rely on data being inputted into Identity variables and/or YesNo variables being clicked on. The failures might be related to the regression I found with those variable types.

There are a couple of other items I'll need to check in the openlaw web app, but those are dependent on the Identity and YesNo variables working correctly.

[jdville03]

This video shows regression with the Period type. It's popping an error after the number is typed.

https://www.useloom.com/share/9f136e4bc8494b2aa802318eb7a6b966

[jtrein]

This is great stuff, I'm going to dig into this and try to reproduce and fix. Thank you.

[jtrein]

@jdville03 OK, I narrowed it down to the way I was handling the cached function for Openlaw.checkValidity. Fixes are pushed.

[jtrein]

@jdville03 OK, className has been replaced for Date that was causing tests to fail in openlaw.

[jdville03]

Still seeing the issues with the YesNo variable after the new fixes: https://github.com/openlawteam/openlaw-elements/pull/46#issuecomment-474597057

[jtrein]

@jdville03 OK, YesNo fixes have been pushed!

[jtrein]

Thank you! Maybe separately, just in case?

[jtrein]

Actually, we don't have any package dependency on openlaw, as it's a prop. We could mention something in the README about what versions we support, which I think right now is anything published in the last few months (guessing here).