core/rawdb/freezer: table size is inaccurate if elements were removed from the tail

[rjl493456442]

Freezer table can support both tail and head deletion. Tail deletion is a bit special, it
will hide the removed items and keep them in the file. The file will only be dropped if
all elements in this file are hidden. (we call it as lazy-deletion).

However, the reported size doesn't respect the tail deletion. The hidden items are
still counted for overall table size. Although these elements are indeed live in the
file system, but we might need another API to report the size of live elements to
avoid some confusions.

[0jovi0]

Can I try it? =)

[shyam-patel-kira]

Is there any update on this? I'd like to try

[theblueskies]

@MariusVanDerWijden Is this still up for grabs? can I get it?

[Sh0g0-1758]

I am also highly interested in contributing to geth. Can I take this one :)

[bnovil]

Does anyone still work on this?

[makinje16]

👋🏿 Just looking to present a different view and to get some clarification since I've started to read through the rawdb code.

From the database perspective, the table size is accurate. The contract when using the freezer is that data is lazily deleted when freezerTables are defined which is also seen on the client side when looking at the AncientWriter interface.

I don't think its a good idea to change this contract because by "deleting" data in the ancient we aren't guaranteeing the data is gone. We guarantee we won't visit it again. I'm all for this use case if we need it but I agree it may want to live somewhere else or under a different method/API/etc.

(I can do this but had this thought while reading through the code)

[wangyifan]

I will give this one a try as my first ticket. It looks like there are multiple PRs in the past but are all abandoned by the author.