Error when using Dagger in Jupyter Notebook / Google Colab

[AkhilAkkapelli]

When attempting to use Dagger in Jupyter Notebook and Google Colab with the following basic commands:

import Pkg; 
Pkg.add("Distributed");using Distributed
addprocs(3)
Pkg.add("Dagger"); using Dagger
Ad = rand(Blocks(50, 50), 100, 100)

I encountered this error sometimes:

UndefVarError: `StreamingFunction` not defined

Stacktrace:
  [1] Dagger.DTaskMetadata(spec::Dagger.DTaskSpec)
    @ Dagger ~/.julia/dev/Dagger/src/submission.jl:223
  [2] eager_spawn(spec::Dagger.DTaskSpec)
    @ Dagger ~/.julia/dev/Dagger/src/submission.jl:233
  [3] spawn(::Any, ::Any, ::Vararg{Any}; kwargs...)
    @ Dagger ~/.julia/packages/Dagger/iu3Pz/src/thunk.jl:487
  [4] spawn(::Any, ::Any, ::Vararg{Any})
    @ Dagger ~/.julia/dev/Dagger/src/thunk.jl:466
  [5] #361
    @ ~/.julia/packages/Dagger/iu3Pz/src/thunk.jl:427 [inlined]
  [6] (::Dagger.var"#361#365"{Dagger.AllocateArray{Float64, 2}})(::Tuple{Int64, ArrayDomain{2, Tuple{UnitRange{Int64}, UnitRange{Int64}}}})
    @ Dagger ./none:0
  [7] iterate
    @ ./generator.jl:47 [inlined]
  [8] collect(itr::Base.Generator{Base.Iterators.Enumerate{Dagger.DomainBlocks{2}}, Dagger.var"#361#365"{Dagger.AllocateArray{Float64, 2}}})
    @ Base ./array.jl:834
  [9] stage(ctx::Context, a::Dagger.AllocateArray{Float64, 2})
    @ Dagger ~/.julia/dev/Dagger/src/array/alloc.jl:40
 [10] _to_darray(x::Dagger.AllocateArray{Float64, 2})
    @ Dagger ~/.julia/dev/Dagger/src/array/darray.jl:92
 [11] rand
    @ ~/.julia/packages/Dagger/iu3Pz/src/array/alloc.jl:50 [inlined]
 [12] rand(::Blocks{2}, ::Int64, ::Int64)
    @ Dagger ~/.julia/dev/Dagger/src/array/alloc.jl:54
 [13] top-level scope
    @ In[29]:1

and other times I get:

MethodError: no method matching DTask(::UInt64, ::Dagger.ThunkFuture, ::MemPool.DRef)

Stacktrace:
  [1] eager_spawn(spec::Dagger.DTaskSpec)
    @ Dagger ~/.julia/packages/Dagger/iu3Pz/src/submission.jl:228
  [2] spawn(::Any, ::Any, ::Vararg{Any}; kwargs...)
    @ Dagger ~/.julia/packages/Dagger/iu3Pz/src/thunk.jl:487
  [3] spawn(::Any, ::Any, ::Vararg{Any})
    @ Dagger ~/.julia/packages/Dagger/iu3Pz/src/thunk.jl:454
  [4] #361
    @ ~/.julia/packages/Dagger/iu3Pz/src/thunk.jl:427 [inlined]
  [5] (::Dagger.var"#361#365"{Dagger.AllocateArray{Float64, 2}})(::Tuple{Int64, ArrayDomain{2, Tuple{UnitRange{Int64}, UnitRange{Int64}}}})
    @ Dagger ./none:0
  [6] iterate
    @ ./generator.jl:47 [inlined]
  [7] collect(itr::Base.Generator{Base.Iterators.Enumerate{Dagger.DomainBlocks{2}}, Dagger.var"#361#365"{Dagger.AllocateArray{Float64, 2}}})
    @ Base ./array.jl:834
  [8] stage(ctx::Context, a::Dagger.AllocateArray{Float64, 2})
    @ Dagger ~/.julia/packages/Dagger/iu3Pz/src/array/alloc.jl:40
  [9] _to_darray(x::Dagger.AllocateArray{Float64, 2})
    @ Dagger ~/.julia/packages/Dagger/iu3Pz/src/array/darray.jl:92
 [10] rand
    @ ~/.julia/packages/Dagger/iu3Pz/src/array/alloc.jl:50 [inlined]
 [11] rand(p::Blocks{2}, dims::Tuple{Int64, Int64})
    @ Dagger ~/.julia/packages/Dagger/iu3Pz/src/array/alloc.jl:55
 [12] top-level scope
    @ In[55]:1

[jpsamaroo]

This is because of addprocs(3) - you're mixing Dagger versions, because your master process is using a different version of Dagger than your worker processes. Using addprocs(3; exeflags="--project=$(Base.active_project())") should fix this.