vigna/dsi-bitstream-rs

A Rust implementation of read/write bit streams supporting several types of instantaneous codes

dsi-bitstream-rs's Issues

Storing the total number of gaps inserted in CodesStats

Hi! Right now CodeStats does not store the number of gaps it has seen. It would be nice if it could store that as metadata so that I would not need to store it externally to compute the percentage of improvement over the naive baseline (storing the hashes themselves). Would it be okay if I added that as a u64?

dsi-bitstream-rs/src/utils/stats.rs

Lines 32 to 55 in 740abcd

	pub struct CodesStats<
	// How many ζ codes to consider.
	const ZETA: usize = 10,
	// How many Golomb codes to consider.
	const GOLOMB: usize = 20,
	// How many Exponential Golomb codes to consider.
	const EXP_GOLOMB: usize = 10,
	// How many Rice codes to consider.
	const RICE: usize = 10,
	// How many Pi and Pi web codes to consider.
	const PI: usize = 10,
	> {
	pub unary: u64,
	pub gamma: u64,
	pub delta: u64,
	pub omega: u64,
	pub vbyte: u64,
	pub zeta: [u64; ZETA],
	pub golomb: [u64; GOLOMB],
	pub exp_golomb: [u64; EXP_GOLOMB],
	pub rice: [u64; RICE],
	pub pi: [u64; PI],
	pub pi_web: [u64; PI],
	}

Use wrap or is_ok uniformly in the docs

make code base clippy-clean

currently we do not pass "cargo clippy" due to the following errors (there are also warnings, but they are not blocking):

error: read amount is not handled
  --> src/backends/file_backend.rs:68:9
   |
68 |         self.file.read(res.as_mut())?;
   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: use `Read::read_exact` instead, or handle partial reads
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#unused_io_amount
   = note: `#[deny(clippy::unused_io_amount)]` on by default

error: written amount is not handled
  --> src/backends/file_backend.rs:79:9
   |
79 |         self.file.write(word.to_ne_bytes().as_ref())?;
   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: use `Write::write_all` instead, or handle partial writes
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#unused_io_amount

set_word_pos does not change position upon error

We should check that PeekType is at most half of Word.

Serde support

Hi! I am running CodesStats to identify which Code may be better for my use case (gaps in a sorted hash list).

I noticed that serde is wholly unsupported, and I would like to have the possibility to dump the Codes obtained by the analysis since I need to run several possible parameterizations. Is there any reason for the lack of serde? Otherwise, I can add that with a pull request.