Julia Data Format - JLD2

JLD2 saves and loads Julia data structures in a format comprising a subset of HDF5, without any dependency on the HDF5 C library. It typically outperforms the JLD package (sometimes by multiple orders of magnitude) and often outperforms Julia's built-in serializer. While other HDF5 implementations supporting HDF5 File Format Specification Version 3.0 (i.e. libhdf5 1.10 or later) should be able to read the files that JLD2 produces, JLD2 is likely to be incapable of reading files created or modified by other HDF5 implementations. JLD2 does not aim to be backwards or forwards compatible with the JLD package.

Reading and writing data

A new interface: jldsave

jldsave makes use of julia's keyword argument syntax to store files, thus leveraging the parser and not having to rely on macros. To use it, write

x = 1
y = 2
z = 42

# The simplest case:
jldsave("example.jld2"; x, y, z)
# it is equivalent to 
jldsave("example.jld2"; x=x, y=y, z=z)

# You can assign new names selectively
jldsave("example.jld2"; x, a=y, z)

# and if you want to confuse your future self and everyone else, do
jldsave("example.jld2"; z=x, x=y, y=z)

Compression and non-default IO types may be set via positional arguments.

save_object and load_object functions

If only a single object needs to stored and loaded from a file, one can use save_object and load_object functions.

save_object(filename, x)

Stores an object x in a new JLD2 file at filename. If a file exists at this path, it will be overwritten.

Since the JLD2 format requires that all objects have a name, the object will be stored as single_sotred_object. If you want to store more than one object, use @save macro, jldopen or the FileIO API.


To save the string hello to the JLD2 file example.jld2:

hello = "world"
save_object("example.jld2", hello)

Returns the only available object from the JLD2 file filename (The stored object name is inconsequential). If the file contains more than one or no objects, the function throws an ArgumentError.

For loading more than one object, use @load macro, jldopen or the FileIO API.


To load the only object from the JLD2 file example.jld2:

hello = "world"
save_object("example.jld2", hello)
hello_loaded = load_object("example.jld2")

save and load functions

The save and load functions, provided by FileIO, provide a mechanism to read and write data from a JLD2 file. To use these functions, you may either write using FileIO or using JLD2. FileIO will determine the correct package automatically.

The save function accepts an AbstractDict yielding the key/value pairs, where the key is a string representing the name of the dataset and the value represents its contents:

using FileIO
save("example.jld2", Dict("hello" => "world", "foo" => :bar))

The save function can also accept the dataset names and contents as arguments:

save("example.jld2", "hello", "world", "foo", :bar)

When using the save function, the file extension must be .jld2, since the extension .jld currently belongs to the previous JLD package.

If called with a filename argument only, the load function loads all datasets from the given file into a Dict:

load("example.jld2") # -> Dict{String,Any}("hello" => "world", "foo" => :bar)

If called with a single dataset name, load returns the contents of that dataset from the file:

load("example.jld2", "hello") # -> "world"

If called with multiple dataset names, load returns the contents of the given datasets as a tuple:

load("example.jld2", "hello", "foo") # -> ("world", :bar)

File interface

It is also possible to interact with JLD2 files using a file-like interface. The jldopen function accepts a file name and an argument specifying how the file should be opened:

using JLD2

f = jldopen("example.jld2", "r")  # open read-only (default)
f = jldopen("example.jld2", "r+") # open read/write, failing if no file exists
f = jldopen("example.jld2", "w")  # open read/write, overwriting existing file
f = jldopen("example.jld2", "a+") # open read/write, preserving contents of existing file or creating a new file

Data can be written to the file using write(f, "name", data) or f["name"] = data, or read from the file using read(f, "name") or f["name"]. When you are done with the file, remember to call close(f).

Like open, jldopen also accepts a function as the first argument, permitting do-block syntax:

jldopen("example.jld2", "w") do file
    file["bigdata"] = randn(5)


It is possible to construct groups within a JLD2 file, which may or may not be useful for organizing your data. You can create groups explicitly:

jldopen("example.jld2", "w") do file
    mygroup = JLD2.Group(file, "mygroup")
    mygroup["mystuff"] = 42

or implicitly, by saving a variable with a name containing slashes as path delimiters:

jldopen("example.jld2", "w") do file
    file["mygroup/mystuff"] = 42
# or save("example.jld2", "mygroup/mystuff", 42)

Both of these examples yield the same group structure, which you can see at the REPL:

julia> file = jldopen("example.jld2", "r")
JLDFile /Users/simon/example.jld2 (read-only)
 └─📂 mygroup
    └─🔢 mystuff

Similarly, you can access groups directly:

jldopen("example.jld2", "r") do file
    @assert file["mygroup"]["mystuff"] == 42

or using slashes as path delimiters:

@assert load("example.jld2", "mygroup/mystuff") == 42

Unpack.jl API

When additionally loading the UnPack.jl package, its @unpack and @pack! macros can be used to quickly save and load data from the file-like interface. Example:

using UnPack
file = jldopen("example.jld2", "w")
x, y = rand(2)

@pack! file = x, y # equivalent to file["x"] = x; file["y"] = y
@unpack x, y = file # equivalent to x = file["x"]; y = file["y"]

The group file_group = Group(file, "mygroup") can be accessed with the same file-like interface as the "full" struct.


Objects are cached during loading

JLD2 caches objects during loading. It may give you the same object twice. This can lead to surprising results if you edit loaded arrays. Note, the underlying file is not being edited!

julia> jldsave("demo.jld2", a=zeros(2))

julia> f = jldopen("demo.jld2")
JLDFile /home/isensee/demo.jld2 (read-only)
 └─🔢 a

julia> a = f["a"] # bind loaded array to name `a`
2-element Vector{Float64}:

julia> a[1] = 42; # editing the underlying array

julia> f["a"]
2-element Vector{Float64}:

julia> a=nothing # remove all references to the loaded array

julia> GC.gc(true) # call GC to remove the cache

julia> f["a"] # a new copy is loaded from the file
2-element Vector{Float64}:


JLD2 tries to write files in a way that allows you to load them on different operating systems and in particular both on 32bit and 64bit systems. However, many julia structs may be inherently different on different architectures making this task impossible. In particular, moving data from a 64bit system to a 32bit system is only guaranteed to work for basic datatypes.


Beware of opening JLD2 files from untrusted sources. A malicious file may execute code on your computer. See e.g. here. To check a file, you can use JLD2DebugTools.jl to view what kinds of objects are stored.