With this new structure, how do I read the attribute value itself?

[denfc]

My problem is not an "issue" as such, but my ignorance. Before the latest update, I loaded the attributes (JLD2.load_attributes(oneFile["Header"])), searched for the first element of the pair to get what I wanted and took its second element to obtain the value.

Now the pairs appear to be gone (or am I no longer looking in the right place?). I can get to the "ReadAttribute", but have not figured out how to get the value, e.g., of "Time" in "Header", out of it:

`
using JLD2
jfile = jldopen("xxx.hdf5")
h = JLD2.get_dataset(jfile, "Header")
h2 = h.attributes # OrderedCollections.OrderedDict{String, Any}
h2S = h2["Time"] # type is JLD2.ReadAttribute

`
Sorry to bother you.

-- dfc

[JonasIsensee]

Hi, no worries. I somewhat intentionally have not yet documented the new api since I might still change it soonish.

You're very welcome to try it and give feedback. (Ask please feel free to ask as many questions as needed)

If you already have a Dataset object, you can use JLD2.attributes(dset)

Otherwise JLD2.load_attributes(g, datasetname) should still work. If it doesn't, I'll fix it. However, all of that API may potentially change in a future release. (Not in a way that breaks old data files, just possibly API functions)

[denfc]

Thanks for the rapid response!

I can use JLD2.load_attributes or JLD2.attributes obtain the array index, but I'll be reading (many) standard hdf5 files that I haven't written, and I'm nervous that something might change in them, and I won't have the value that I think I have. With the pairs I looped through, with a "findfirst", until I found what I wanted, and then I took its value.

I can always switch back to the HDF5 package in the interim, but I really like using JLD2.

Thanks again for the info. I work pretty much by myself, so I go down many wrong paths, and a little direction is always greatly appreciated.

-- Denis

[JonasIsensee]

Findfirst should still be ok.

You got an OrderedDict. The entries are guaranteed to be in the correct order und iteration or e.g. (keys(dict) values(dict)

[denfc]

The values are "ReadAttribute". Can I get the actual number easily? For example,

`
julia> h2["Time"]
JLD2.ReadAttribute(:Time, JLD2.ReadDataspace(0xff, 0x00, 6000), JLD2.FloatingPointDatatype(0x11, 0x20, 0x3f, 0x00, 0x00000008, 0x0000, 0x0040, 0x34, 0x0b, 0x00, 0x34, 0x000003ff), 6000)

`

The actual number for "Time" is 0.33300950325867074.

[JonasIsensee]

Ah, damn.

For now, use:
JLD2.read_attr_data(f::JLDFile, attr::ReadAttribute)

I'll fix it, next time I'm at a computer...

[denfc]

Ahh! Great, thanks. I had searched through the "read" functions but either missed that one or applied it incorrectly. Thanks again for the fast assistance! I can now continue to confront my ignorance of the non-Julia issues for which I'm using Julia. Many, many thanks.

[JonasIsensee]

Using the dev branch should make JLD2.attributes(dset::Dataset) and JLD2.load_attributes(::JLDFile, ::String) return an OrderedDict with keys and the actual values in the order they are found in the file.

[denfc]

Maybe not yet? Yesterday's fix still works, so I'm fine, but I get the same result now with the dev version and the non-dev version.

With the dev: [033835bb] JLD2 v0.4.51 ~/.julia/dev/JLD2

using JLD2
jfile = jldopen("xxx.hdf5")
h = JLD2.get_dataset(jfile, "Header")
oDict = JLD2.load_attributes(jfile, "Header")
o = JLD2.attributes(h)
println(o == oDict)
println(typeof(o))
close(jfile)

┌ Warning: This file was not written with JLD2. Some things may not work.
└ @ JLD2 ~/.julia/packages/JLD2/bfPAo/src/JLD2.jl:283
true
Vector{Any}

[denfc]

I appreciate your persistence. AFK for about another week but will eagerly work with it when I return home. Thanks! — Denis

…

On Tue, Jul 30, 2024 at 6:10 PM JonasIsensee ***@***.***> wrote: #580 <#580> This PR should make JLD2.attributes(dset::Dataset) and JLD2.load_attributes(::JLDFile, ::String) return an OrderedDict with keys and the *actual* values in the order they are found in the file. — Reply to this email directly, view it on GitHub <#579 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AE2LUY3LDUYEI2A3EG5FI2LZO7CJ7AVCNFSM6AAAAABLUQTJBWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENJYHAZDKOJRG4> . You are receiving this because you authored the thread.Message ID: ***@***.***>