typemap has no effect in CSV.read in Jupyter Notebook

[shuofanzhang]

CSV.read(HTTP.get("https://raw.githubusercontent.com/shuofanzhang/covid/main/Test_pos_over_time.csv").body,
DataFrame;
footerskip=2,
select = [1, 2, 6],
typemap = Dict(Float64 => String),
dateformat= "yyyy-mm-dd")

Try above code in Jupyter Notebook (6.4.8) with CSV v0.8.5, the resulting columns are not String but Float64.

But the same codes works as expected in VS Code.

[nickrobinson251]

most likely in the VS Code environment a newer version of CSV.jl is being used. The latest version is v0.10. At this point v0.8 is relatively out-dated and in general we do not backport bugfixes to old versions. Is it possible for you to update to CSV.jl v0.10?

[shuofanzhang]

I see, I just downloaded the newest Julia on my new laptop, so I thought everything was the latest.

It seems like I cannot update the CSV package, I get this information when I do status --outdated

But I am not sure how to fix it...

[quinnj]

It's not the clearest, but in the yellow (green?) text after CSV v0.8.5 it says Forecast, so that tells you the package that is "holding you back". I've made a PR here to update that package to be compatible w/ 1.0. Once we get CSV.jl 1.0 tagged, that should help these kinds of packages stay more up to date since they can just put compat at CSV: "1" and be done for a long time.