Partitioned phf

[rickbeeloo]

I think we can already add this.

The lookups were tested with https://github.com/rickbeeloo/UniqueMers, with 2^32 unique 17-mer lookup being correct according to the sshash --check (UniqueMers kmers.fasta 17 4294967296 0).

I just added "tail" to the linked repo to add random sequences after each kmer allowing for the navigational queries to be tested, but instead of re-running sshash again for this I think we can better do Fulgor immediately.

[rickbeeloo]

Hey @jermp, what did you actually mean? The lookup struct right (the contig size)?

[jermp]

Hey @jermp, what did you actually mean? The lookup struct right (the contig size)?

Yes, I would default everything to uint64_t.

[rickbeeloo]

There we go, everything to uint_64:

• Added msb_uint64, removed msb (uint32)
• Removed the constant invalid_uint32 mainly cause Fulgor reads these from sshash . I just read the uint64 constant in my branch, but to be sure output from this branch isn't accidently verified using the uint32 constant.
• The only thing we still have uint32 is the crc for the gzips

Let me know :)

P.s. I think you should do a read haha cause for example I removed this pragma push as it's now two uint64s but maybe it's necessary in c++ to have it for 8 bytes here.

[jermp]

Sorry for the confusion but in my comment I was referring to the uint32_t inside the lookup_result struct. The other place where uint32_t is used are just fine. So I'm gonna close this pull request and make the change myself to avoid conflicts. Thanks!

[jermp]

Date as of https://github.com/jermp/sshash/commit/fbe00d9f1ca4fa875177deba11a7e9b9e5c2b71a.