CodeTracking does not find function unless in another package?

[co1emi11er2]

I have this package I made called Handcalcs which uses this package to "unroll" functions. See here. This has been pretty great to use although I am hoping to get it working for when functions are defined in the current scope.

Say I have this script:

using CodeTracking

function my_add(a, b)
    c = a + b
end

f = @code_expr my_add(3, 4)

println(f)

d = @which my_add(3, 4)

println(d)

The output I get is:

nothing 
my_add(a, b) @ Main c:\Users\c-mil\jlcode\codetracking\codetrack.jl:3

So the @code_expr macro fails to find the function just defined, but the @which macro is able to. Is there a way to get this working? I could try and implement it if you think there is a way. I think this could also fix the jupyter issue (#51) as well?

[timholy]

@code_expr requires Revise to work: https://github.com/timholy/CodeTracking.jl?tab=readme-ov-file#code_string-and-code_expr
Your example works for me if I load Revise first.

Fixing #51 would be great though!

[co1emi11er2]

Hmm.. Are you sure you didn't shift+enter into the REPL in some way? I still get this when running the script as a script.

using CodeTracking
using Revise

function my_add(a, b)
    c = a + b
end

f = @code_expr my_add(3, 4)

println(f)

d = @which my_add(3, 4)

println(d)

The output I get is still:

nothing 
my_add(a, b) @ Main c:\Users\c-mil\jlcode\codetracking\codetrack.jl:8

[co1emi11er2]

I have to add the function into the code like so to get it to work:

using CodeTracking
using TestHandcalcFunctions

using Revise

isdefined(Main, :Revise) ? Main.Revise.includet("script.jl") : include("script.jl")

f = @code_expr my_add(3, 4)

println(f)

d = @which my_add(3, 4)

println(d)

Where script.jl contains the my_add function. But yeah it is really to solve the jupyter notebook issue and also Pluto notebooks as well. You can leave this closed since the other issue is still open.

Thanks for the package. It is allowing me to do things I don't think is really possible in another language.

[timholy]

For me I just copy/paste your original example. Are you running in the REPL or is this IJulia?

[timholy]

Let's reopen this since I'm guessing this is IJulia-related. If so, can you add IJulia to the title of this issue?

[co1emi11er2]

This is in VSCode. I have a codetrack.jl script and I am running it by clicking the run button. I can also just open a REPL and run include("codetrack.jl") and it will behave the same way.

[timholy]

Hmm, OK. Does it work for you at the REPL if you copy/paste?

[co1emi11er2]

Yes it will work that way.

[timholy]

It kind of makes sense that include("codetrack.jl") doesn't work. CodeTracking needs a place to "file" the defining expressions, and it goes to some effort to make that work for the REPL. includet("codetrack.jl") should also work, but not if you're still in the middle of evaluating the toplevel lowered code block:

tim@hypnotic:/tmp/ct$ julia -q --project
julia> include("codetrack.jl")                        # first time with plain `include`
nothing
my_add(a, b) @ Main /tmp/ct/codetrack.jl:4

julia> @code_expr my_add(3, 4)

julia>
tim@hypnotic:/tmp/ct$ julia -q --project
julia> includet("codetrack.jl")                       # now with `includet`
nothing
my_add(a, b) @ Main /tmp/ct/codetrack.jl:4

julia> @code_expr my_add(3, 4)
:(function my_add(a, b)
      #= /tmp/ct/codetrack.jl:4 =#
      #= /tmp/ct/codetrack.jl:5 =#
      c = a + b
  end)

Basically, Revise hasn't had a chance to cache it before you're asking for it.

[timholy]

When you split all these lines up at the REPL, Revise has already completed its work and it's available the first time you ask for it.

[co1emi11er2]

Yeah. The good thing is in IJulia, there is a workaround. I can just call includet on a file with defined functions and things work as expected. It just requires a little learning curve for new users.

[timholy]

Anything left to do here? If not, let's close it. Bottom line: Revise must be loaded and you have to wait to access the code via @code_expr until after Revise has had a chance to cache it.

[co1emi11er2]

No. We can close it