This is a Swift wrapper for the HDF5 file format. HDF5 is used in the scientific comunity for managing large volumes of data. The objective is to make it easy to read and write HDF5 files from Swift, including playgrounds.
This example shows how to open an existing HDF5 file and write data to an existing dataset.
import HDF5Kit
// Initialize the data
let dataWidth = 6
let dataHeight = 4
var data = [Double](repeating: 0.0, count: dataHeight * dataWidth)
for r in 0..<dataHeight {
for c in 0..<dataWidth {
data[r * dataWidth + c] = Double(r * dataWidth + c + 1)
}
}
// Open an existing file
let path = "file.h5"
guard let file = File.open(path, mode: .readWrite) else {
fatalError("Failed to open \(path)")
}
// Open an existing dataset
let datasetName = "dset"
guard let dataset = file.openDoubleDataset(datasetName) else {
fatalError("Failed to open dataset \(datasetName)")
}
// Write the data
try dataset.write(data)Reading data is really easy with HDF5Kit:
// Open an existing file
let path = "file.h5"
guard let file = File.open(path, mode: .readWrite) else {
fatalError("Failed to open \(path)")
}
// Open an existing dataset
let datasetName = "dset"
guard let dataset = file.openStringDataset(datasetName) else {
fatalError("Failed to open dataset \(datasetName)")
}
let data = dataset[1...3, 2...5]Supported types are: Double, Float, Int and String.
HDF5Kit has been updated for Swift 6 with several powerful new features and architectural improvements.
We have consolidated the type-specific classes into a unified generic system. While IntDataset, DoubleDataset, etc., are still available as typealiases for backward compatibility, you can now use the generic Dataset<T> and Attribute<T> where T: HDF5Representable.
Attributes can now be attached to any HDF5 object, including both Groups and Datasets.
let dataset = file.createIntDataset("results", dataspace: space)!
let attr = dataset.createStringAttribute("units")!
try attr.write("meters")You can now easily enable zlib (deflate) compression when creating datasets by passing an optional compression level (0-9).
// Create a compressed dataset (deflate level 9)
let dataset = try file.createAndWriteDataset("compressed_data",
dims: [1000, 1000],
data: largeData,
compression: 9)Support for Soft Links and External Links is now available.
// Create a soft link (alias) inside the file
try file.createSoftLink(targetPath: "/very/long/path/to/data",
linkName: "shortcut")
if file.linkExists("shortcut") {
let dataset: DoubleDataset? = file.openDoubleDataset("shortcut")
}- @HDF5Actor: All C-API calls are now serialized via a global actor to ensure thread safety with the non-thread-safe libhdf5.
- Safe Raw Handle Access: Advanced C interop should use
withUnsafeID { id in ... }so raw HDF5 identifiers are only accessed while isolated to@HDF5Actor. The identifier is owned by HDF5Kit and must not be stored after the closure returns. - Improved Error Handling: Captures descriptive HDF5 error messages for better debugging.
- Modern Pointers: Fully updated to use Swift 6 pointer paradigms.
let objectName = await dataset.withUnsafeID { datasetID in
// Call lower-level HDF5 C APIs here while serialized by HDF5Kit.
String(describing: datasetID)
}