Closed nicolamos closed 3 weeks ago
UUIDs are great for completely unambiguous references, but I completely get that from a usability point, they're not great. That's why dataset
supports the syntax dataset("name", "key" => value...)
.
For example, with this setup
data_config_version = 0
uuid = "663bbc0d-575d-4caf-ab51-b2ec6737a15b"
name = "demo"
[[atest]]
uuid = "92b63dcf-949f-4439-af84-f20996166481"
date = 2024-04-01
[[atest.storage]]
driver = "raw"
value = 7
[[atest.loader]]
driver = "passthrough"
[[atest]]
uuid = "17db2f80-8f48-4344-8f09-894c6676331f"
date = 2024-04-07
[[atest.storage]]
driver = "raw"
value = 26
[[atest.loader]]
driver = "passthrough"
You can use dataset
like so:
julia> dataset("atest", "date" => Date("2024-04-01")) |> read
7
julia> dataset("atest", "date" => Date("2024-04-07")) |> read
26
Let me know how that seems to you :slightly_smiling_face:
In the upcoming v0.10, I'll also be showing more information about the dataset in the REPL:
Between this and my earlier comment, it should be pretty easy to identify and get a datasets based on its parameters.
Thank you for the updates. Having the parameters available shown in the REPL is very useful.
When defining multiple datasets with the same name, seems very hard to retrieve them by specifying eg. the date. Using UUIDs for that seems inconvenient as I think it is easier to specify parameters that are actually present in the data.