@romgrk/hdf5
v0.3.8
Published
Safer native API for HDF5 files
Downloads
1
Readme
This if a fork of the official hdf5 module, with more complete bindings and less foot-shooting capabilities.
Documentation
API documentation is available at http://hdf-ni.github.io/hdf5.node.
Differences with the official documentation
fileOrGroup.setAttribute(name, value, options?)
To set attributes, the .setAttribute method is added to give more options.
| argument | type | comment |
|-----|-----|-----|
|name|string| Name of the attribute |
|value|any| Value of the attribute |
|options?|{ padding: Number }| If value is a string, sets the string-padding of the type |
file.close()
Will throw an error if there are still open groups, as it would cause bugs further down the road not to close them.
h5lt.makeDataset(parentId, name, values, options)
This method is extended to allow creation of more complex string-types, with ranks > 1.
| argument | type | comment |
|-----|-----|-----|
|parentId|Number| Id of the parent group or file |
|name|string| Name of the dataset |
|values|Array|Buffer| Values |
|options?|Object| Optional parameters |
|options.fixedSize|boolean| If values is an array of string, will be stored as fixed-size strings |
|options.padding|Number| If values is an array of string, sets the string-padding of the type |
(Only the differences with the official doc are listed here)
Installation
npm install hdf5See http://hdf-ni.github.io/hdf5.node/doc/install-setup.html for the native requirements and details. If your native hdf5 libraries aren't at the default you can set the path with --hdf5_home_linux switch on this project as well as dependent projects.
npm install hdf5 --hdf5_home_linux=<your native hdf path>For mac and windows the switches are --hdf5_home_mac & --hdf5_home_win
To install with yarn first need to configure so it knows where the libraries are:
yarn config set hdf5_home_linux $HDF5_HOME
yarn installNote: If node-gyp isn't installed
npm install -g node-gypQuick start to open and read from an h5 file
var hdf5 = require('hdf5').hdf5;
var Access = require('hdf5/lib/globals').Access;
var file = new hdf5.File('/tmp/foo.h5', Access.ACC_RDONLY);
var group = file.openGroup('bar');Notes on Recent Releases
Note: Release v0.3.5
- Strictly for upgrading the build on Windows. Tested with VS 2017
Note: Release v0.3.4
- Reference attributes on datasets and groups are now available properties.
- Reserved propeties such as type, rank, rows etc. are now settable in options for dataset functions.
- Typescript definition files now available.
- For static native linkng, link_type command line switch is provided in binding.gyp(darwin,win untested).
- Added custom 64 bit signed (Int64) and unsigned (Uint64) integer attributes read/write since they aren't yet supported by javascript.
- Added a file method enableSingleWriteMultiRead (if native version older than 1.10.x it is a noop).
- Synchronous iterate and visit is now available on file and group children.
- Bug fixes on dimensioning have been made
Note: asynchronous i/o is coming but not in this release
Note: Release v0.3.3 Minor fix on fixed length of strings in array. Handles the situation where the strings are contiguous without all having null bytes.
Note: Release v0.3.2 tested with nodejs v6.11.2, v7.5.0 and v8.4.0. Code was changed to allow v8.4.0 to work while still working with the earlier versions. It may work back as far as v4.2.1 let me know if you have a version in between that needs testing. Variable length array of strings can be be read as regions.
Note: Release v0.3.1 is based on new V8 API changes coming with nodejs v7.
Note: Release v0.1.0 was built with nodejs v4.2.1. If you want nodejs v0.12.x then stay with release v0.0.20. npm will continue with nodejs v4.x.x line and any fixes or features needed by prior versions will be from github branches.
Note: Release v0.0.20 is for prebuilts with hdf5-1.8.15-patch1. If you want hdf5-1.8.14 stay with v0.0.19.
Philosophy
This module, hdf5.node, is intended to be a pure API for reading and writing HDF5 files. Graphical user interfaces or other layers should be implemented in separate modules.
Unlike other languages that wrap hdf5 API's this interface takes advantage of the compatibility of V8 and HDF5. The result is a direct map to javascript behavior with the least amount of data copying and coding tasks for the user. Hopefully you won't need to write yet another layer in your code to accomplish your goals.
Other Feature Notes
The node::Buffer and streams are being investigated so native hdf5 data's only destination is client browser window or client in general.
Dimension Scales
Mostly implemented (missing H5DSiterate_scales[ found a way to make callback functions from te native side and looking to finish this and use the technique for other h5 iterators])
High-level Functions for Region References, Hyperslabs, and Bit-fields
Writing an interface based on the standard hdf5 library. Currently you can write and read a subset from a two rank dataset. Other ranks may work yet are untested. See tutorial http://hdf-ni.github.io/hdf5.node/tut/subset_tutorial.html for example applied to node Buffers.
Filters and Compression
Testing filters and compression. Have the gzip filter working. For some applications getting the uncompressed data from the h5 would reduce the number of compressions and decompressions. For example an image could be sent to client before unzipping and rezipping on the server side.
Third party filters can be used. Those do take separate compiled libraries yet are independent. They get picked up by native hdf5 from the HDF5_PLUGIN_PATH.
Image
The h5im namespace is being designed to meet the Image Spec 1.2 http://www.hdfgroup.org/HDF5/doc/ADGuide/ImageSpec.html. Hyperslabs/regions of images can now be read.