Consider excluding the patch version & `-(RC|M)\d+` from indexing

[rwinch]

We should consider excluding the patch version, milestone, and RC from the version in the search.

Steps

Helper Script

I believe to solve this we should add a helper script that strips out the patch, milestone, and RC information form the version. It should not strip out the -SNAPSHOT since that is something that needs qualified on.

For example, the versions would output:

• 1.0.0 -> 1.0
• 1.0.0-SNAPSHOT-> 1.0-SNAPSHOT`
• 1.0.0-M1 -> 1.0
• 1.0.0-RC1 -> 1.0

Update Templates

Then it would be placed in the meta version tag and data-page-version attributes here and here.

Results

Initial Upgrade

When projects switched to this template they would experience about 24h of search not working. Once the search crawled the page with new UI, it would then be indexed without the patch and on release day it would just point to the previous release (assuming there was one).

Going Forward

When releases happened, there might be some stale data (deltas between the two pages) but it wouldn't have no results in most cases. The exception is that there would be no results for the first milestone or first SNAPSHOT of a <major>.<minor> until the next crawl happened.