Corrosiff

A Rust-based reader for .siff file data. An excuse to learn and use Rust, which I have to admit has been pretty enjoyable.

While most functionality is .siff-exclusive (because it's built around arrival time data), it can read frames from .tiff files too.

TODOS:

• A real README.

• Testing has gotten reaaaaaal lax...

• More methods for estimating "lifetime" (e.g. RLD, MLE...)

• mask methods for tau_d or for .tiff files.

• Fast timestamp checking without reading as much metadata? Make a scan_timestamps method.

• get_epoch_timestamps_both doesn't error if system timestamps don't exist! It will just crash! Because I don't use the ? correctly.

• C-compatible FFI.

• More sophisticated macros so that I don't have to manually go through each file to make every combination of registered / unregistered / mask / masks + flim. This seems like exactly the type of thing the Rust macro system is perfect for (once I understand it better!). This is obviously a very bad boilerplate-y v0.0.1 type implementation.

It also seems like this is a very natural place for using better structs and traits. I think what I want to do is create a trait for all the varied return types,

trait FileReadImage <D : Dim> {
    pub fn read_raw(...) -> Result<(), IOError>;
    pub fn read_compressed(...)
    pub fn read_raw_registered(...)
    pub fn read_compressed_registered(...)
}

struct MaskArray{}

impl FileReadImage for MaskArray {
    ...
}

for things like sum_mask, sum_masks, etc. The tricky part is that each one needs some subtly different args...

• Consider making more useful and interesting Rust structs for return values, rather than just passing back ndarray objects. I could wrap these in, for example, the SiffFrame struct buried deep in the data submodule and then would have access to some faster operations implemented in Rust rather than relying on my Python interface to do everything complicated (though access to numpy really is nice...)

Installation

---

This is not yet a publicly available crate, so for now you have to download it and install it yourself!

First clone the repository:

git clone [email protected]:MaimonLab/Corrosiff.git

then navigate to the directory containing the repository. You can then build it with cargo build --release, providing libraries for use by other Rust code (such as the corrosiffpy package).

Alternatively, you can simply specify this repository in your Cargo.toml:

[dependencies]
corrosiff = { git = "https://github.com/MaimonLab/Corrosiff.git" }

Sections

---

File data

Image Time

Image

Metadata

Troubleshooting

---

Haven't had any problems yet, so I'm not sure what to put here!

Testing

---

This is not really resolved -- right now I run all the tests on local files from my computer! I'm working on hosting remote files to download and test automatically! I'm just still learning this stuff..

The main modules have several tests built in. From the main directory corrosiff, run

cargo test

and it will run the test suite and print the results.

If any of the tests fail, let me (SCT) know and I'll do my best to address the problems.

Benchmarking

The corrosiff library was implemented to make up for my poor C/C++ skills. Some of the siffreadermodule calls of the Python extension module were slow -- much slower than I'd like. So I decided I'd learn Rust and see how that changes.

This section documents the speed on the computers I've tested so far.