Implement of searchv2 to fetch batch of blocks

[mike-petrov]

closes #34

Signed-off-by: Mikhail Petrov mike@nspcc.ru

[mike-petrov]

While tetsing, I found an interesting case, some blocks have a duplicate:

[roman-khimov]

It's known, there are duplicates. Code should handle this.

[mike-petrov]

@roman-khimov everything works now, asynchronously, but because of duplicates I can't understand from which number I should request a search:

i.e. if I specify the interval 1000-2700, then we take first 1000 objects with 50 duplicates, and the second search request will start from 2000, while it should start from 1950, how to solve this situation?

[roman-khimov]

By using the answer you have from the first request? What's the problem, really? You know the last block-oid pair after every request, you don't need to have perfect 1000 blocks.

[mike-petrov]

Yes, I have 1000 blocks every time, but if there is a duplicate among them, I will not know about it at the asynchronous search stage and will skip it:

• first searchv2 (get 0-1000 blocks)
• iterating 0-500 block (get data)
• second searchv2 (get 1000-2000 blocks)
• iterating 500-1000 block (get data, but the 800 number is a duplicate, so we're skipping it)

At this point we missed block number 1000 because of the duplicate

• iterating 1000-1500 block (get data)
  ...

[roman-khimov]

second searchv2 (get 1000-2000 blocks)

get 995-1990, what's the problem? You know where to start from right after the first search.

[mike-petrov]

If I immediately request the last block, then yes - I'll know which number is the last, is that what we want to do?

[mike-petrov]

ok, now it works