nimhdf5

nimhdf5 copied to clipboard

Windows support

1

It seems like `nimhdf5` on Windows might be broken entirely. Simply opening a file seems to lead to a segfault.

Vindaar

fix issues high level example

1

While fixing the high level examples in #43, I noticed that there are multiple regressions in it. Go back and fix them!

Vindaar

Generate docs

1

Generate docs for the module.

Vindaar

Opening file if already opened elsewhere can lead to weird runtime errors

1

If a file is opened in one place and then opened in another place, certain HDF5 functions can fail with some really hard to debug errors. We should try to...

Vindaar

`H5FD_CORE` is a file driver that is part of the HDF5 library, which creates files purely in memory. This can be useful to e.g. create a file and send it...

Vindaar

add iterators for groups, attributes

2

Have some file aware iterators for groups and attributes. While groups are stored in a table, the current way to iterate over the keys of the table sort of works,...

Vindaar

[Design] change storage of groups / datasets

1

The storage of datasets and groups within a `H5FileObj` and `H5Group` should be redone. First of all the objects should be replaced by `ref` objects, because that would make all...

Vindaar

now reading from that file, dataset dimensions seem to be mismatching

3

Error: unhandled exception: Wrong input shape of buffer to write to in `read`. Buffer shape `@[1497098, 2000]`, dataset has shape `@[1497098, 2000]` [ValueError]

clive-g-brown

Setup travis CI

1

Setup travis CI. We should make sure to test all supported HDF5 versions. See `h5py` travis setup and especially the script to download the HDF5 source: https://github.com/h5py/h5py/blob/master/ci/travis/get_hdf5_if_needed.sh

Vindaar

Currently the procs to read a subset from a dataset instead of the whole will broadcast the `seq[int]` indices given to all dimensions of an N dimensional dataset. This was...

Vindaar