Precompute list of needed blobs before a checkout

[jrbriggs]

Rather than fall back on read-object one-by-one, let's precompute what's needed, a la partial.

[derrickstolee]

For discussion: To get this ready for the Demo milestone, we may want an in-between:

1. I'll get the sparse-checkout builtin completed, which performs the read-tree and reset operations.

2. We can create a scalar sparse-checkout verb that first runs a scalar prefetch --folders then runs git sparse-checkout add.

Long-term, we definitely want Git to handle this by itself. However, to get this full list precomputed we will need Git changes for the precompute (this issue) and the batched read-object hook (#7).

[jrbriggs]

@derrickstolee Thanks for starting the discussion.

I think your proposal satisfies the requirements for manipulating the set of folders in the sparse enlistment without moving what HEAD points to, but I don't think it fully addresses computing the list of blobs when we checkout or reset. Right now scalar prefetch --folders reads HEAD to determine the commit ID to prefetch, so I think there's a bit of extra work in step 2 if we also want this to address moving the HEAD ref.

[derrickstolee]

Right now scalar prefetch --folders reads HEAD to determine the commit ID to prefetch, so I think there's a bit of extra work in step 2 if we also want this to address moving the HEAD ref.

True. My proposal will only fix that initial hydration on a fresh clone. A branch-switch would still use the read-object hook for the blobs that changed between.

[github-actions]

Labeling this issue as stale. There has been no activity for 30 days. Remove stale label or comment or this issue will be closed in 7 days.

[github-actions]

Labeling this issue as stale. There has been no activity for 30 days. Remove stale label or comment or this issue will be closed in 7 days.

[jeffhostetler]

This is being addressed in: microsoft/git#191

[jeffhostetler]

marking closed along with #7