Data Docs build does not build suite pages if being run via a checkpoint and does not upload assets

[Aylr]

Describe the bug
Data Docs build does not build suite pages if using a GCS bucket

To Reproduce
Steps to reproduce the behavior:

1. Create a number of suites
2. Set up a GCS site
3. great_expectations docs build
4. Note a GCS site with an index page only - no other files or assets.

Expected behavior
I expect a full datadocs site

Environment (please complete the following information):

• OS: [e.g. iOS] macOS
• GE Version: [e.g. 0.10.0] 0.12.3

[Aylr]

[anthonyburdi]

FYI while trying to reproduce this: currently great_expectations docs build seems to work as expected, however this issue does crop up when using great_expectations checkpoint run ...

[anthonyburdi]

FYI this issue was reproduced on s3 & Filesystem as well when using great_expectations checkpoint run <checkpoint_name>

[anthonyburdi]

Quick update @Aylr @eugmandel - I have a proposed method to fix this in branch anthony/bugfix/issue_1944 and am adding tests before submitting a PR.
The proposed fix is to check for the existence of a data docs site (via a method on HtmlSiteStore to check for an existing index.html) and if it does not exist, recreating the data docs site from scratch during an SiteBuilder.build() by removing the resource_identifiers. This will keep existing functionality as-is, i.e. running a checkpoint for an existing data docs site will only add a page for the new validation and fix the linked issue by rebuilding data docs in full if they don't exist.
Feedback / other ideas welcome!

[Aylr]

A note of mild warning - I vaguely recall the original intent of the resource identifier being passed was to build only the new page rather than the very slow rebuild of the entire site. If a user has many suites (or few large ones), or many validation results this can take quite some time to do.

[anthonyburdi]

Makes perfect sense! Just to be clear - the resource_identifier is only removed if there is no existing site, in which case you would likely want to rebuild. However noted that it could take a very long time, and a user may not have set a validation_results_limit.

Maybe there is a middle ground where if there is no existing site during a checkpoint run, we just make sure the index / expectation / static sections of the site are built (in addition to the given resource identifier for the validation). I'm looking into DefaultSiteSectionBuilder.build() to see if we can force a build of the expectations pages if the site does not exist and the given resource_identifier is not an ExpectationSuiteIdentifier (assuming that in this case a user would just want to render that single expectation suite).

[anthonyburdi]

@Aylr FYI - I changed the functionality to the following: All expectation suites are always rendered unless resource_identifiers contains ExpectationSuiteIdentifier(s). If resource_identifiers contains ExpectationSuiteIdentifier(s) then only those are rendered. PR is now ready for review.