Starting the kernel with IPython 4.0 gives the following error:

ihaskell: main/Main.hs:(66,1)-(73,35): Non-exhaustive patterns in function ihaskell

When run from IPython (4.0) notebook, the kernel is reported as dead but seems to execute cells perfectly useably.

When I run IHaskell notebook I can see tips and suggestions on functions signatures or evaluate :info command, but I can't evaluate cells. It just swallows them silently. When running IHaskell console it's even worse, it freezes in cell In [1]: and trying to kill it with Ctrl+C brings again an input prompt of the console. I am using Ubuntu 12.04. How can I figure out what's going wrong?

I'm trying to create a MacPorts build of IHaskell here: https://github.com/macports/macports-ports/pull/4706

The ihaskell binary produced by stack install does not appear to be independent of the .stack directory, as I believe it should be.

When I launch jupyter notebook and initiate a Haskell kernel, stack jumps in and starts downloading ghc, and the kernel fails.

Is it possible to produce a binary independent of the .stack directory?

The build is produced with:

stack install --system-ghc …

and the file /opt/local/share/jupyter/kernels/haskell/kernel.json says:

{"display_name":"Haskell","argv":["/opt/local/bin/ihaskell","kernel","{connection_file}","--ghclib","/opt/local/lib/ghc-8.6.5","+RTS","-M3g","-N2","-RTS","--stack"],"language":"haskell"}

The broader reason for this is that all MacPorts installs are done through a destroot, and destrooting a common .stack for many independent stack projects won't work.

The strategy that works for other stack projects is to build the binary with a temporary .stack, then throw it away after the working binary is created.

This strategy isn't working with IHaskell for the reasons above.

I installed ihaskell completely from scratch using stack LTS 5.2 on two different machines running debian x86_64:

stack install --no-system-ghc ihaskell && stack exec --no-system-ghc ihaskell -- install --stack && stack exec --no-system-ghc jupyter -- notebook --debug --no-browser --ip=<IP address>

I encountered the same error in both cases:

ihaskell: /home/peddie/.stack/snapshots/x86_64-linux/lts-5.2/7.10.3/lib/x86_64-linux-ghc-7.10.3/cryptonite-0.10-9z0j8QI27Av2VIWw0mEkTO/libHScryptonite-0.10-9z0j8QI27Av2VIWw0mEkTO.a: unhandled ELF relocation(RelA) type 42

ihaskell: ihaskell: unable to load package `cryptonite-0.10'
[I 22:50:20.392 NotebookApp] KernelRestarter: restarting kernel (1/5)

stack exec --no-system-ghc ghc-pkg -- list has no trouble locating cryptonite. stack exec --no-system-ghc ghci has no trouble loading and running functions from Crypto.Random.EntropyPool (to choose a random example).

My kernel.json file:

[email protected]:~$ cat /home/peddie/.local/share/jupyter/kernels/haskell/kernel.json
{"display_name":"Haskell","argv":["/home/peddie/.stack/snapshots/x86_64-linux/lts-5.2/7.10.3/bin/ihaskell","kernel","{connection_file}","--ghclib","/home/peddie/.stack/programs/x86_64-linux/ghc-7.10.3/lib/ghc-7.10.3/","--stack"],"language":"haskell"}

This file was created with /opt/ghc/7.10.3/lib/ghc-7.10.3/ in the --ghclib field of kernel.json, but the error did not go away when I manually replaced it with the path to stack's libs.

EDIT: I forgot a critical piece of information: this issue does not occur when building with stackage LTS 5.1, which also uses cryptonite-0.10.

I recently switched from using Haskell Plarform to using Homebrew for managing my Haskell installation. This change was without issues for GHCi but IHaskell is now unable to find many packages. For example if I

import Data.List.Split 

I get

Failed to load interface for ‘Data.List.Split’

even though the same import statement works fine in GHCi.

Is there something I can do to fix this? What would be causing IHaskell to be looking in the wrong location for installed packages?

FWIW, this occurred following an upgrade from OS X 10.10 to 10.11.

when running cabal install ihaskell I get the following error regardless of which zeromq installation procedure I use (homebrew and github):

Resolving dependencies...
Configuring zeromq4-haskell-0.6.1...
cabal: The pkg-config package libzmq version >=4.0 && <5.0 is required but it
could not be found.
Failed to install zeromq4-haskell-0.6.1
cabal: Error: some packages failed to install:
ihaskell-0.4.2.0 depends on zeromq4-haskell-0.6.1 which failed to install.
zeromq4-haskell-0.6.1 failed during the configure step. The exception was:
ExitFailure 1

The path to my libzmq folder is in ~/.cabal/bin/libzmq.

I may just be doing this wrong, but when a code cell evaluates to something that should be displayed nicely (like a JuicyPixels Image), it just says

Unshowable:Image Pixel8
No instance for (Show (Image Pixel8)) arising from a use of ‘print’
In a stmt of an interactive GHCi command: print it

I already did cabal install ihaskell-juicypixels.

I installed haskell platform through homebrew I ran:

pip3 install jupiter
git clone http://www.github.com/gibiansky/IHaskell
cd IHaskell
./macos-install.sh

Jupiter works but IHaskell errors

Downloading the latest package list from hackage.haskell.org
Skipping download: Local and remote files match.
Resolving dependencies...
Downloading alex-3.1.7...
Downloading happy-1.19.5...
Configuring alex-3.1.7...
Configuring happy-1.19.5...
Building happy-1.19.5...
Building alex-3.1.7...
Failed to install happy-1.19.5
Build log ( /Users/<User>/.cabal/logs/happy-1.19.5.log ):
[1 of 1] Compiling Main             ( /var/folders/_m/1h8dd_fd3dx3qvtq87_c4q480000gn/T/cabal-tmp-85242/happy-1.19.5/dist/setup/setup.hs, /var/folders/_m/1h8dd_fd3dx3qvtq87_c4q480000gn/T/cabal-tmp-85242/happy-1.19.5/dist/setup/Main.o )
Linking /var/folders/_m/1h8dd_fd3dx3qvtq87_c4q480000gn/T/cabal-tmp-85242/happy-1.19.5/dist/setup/setup ...
Configuring happy-1.19.5...
Building happy-1.19.5...
Preprocessing executable 'happy' for happy-1.19.5...
<command line>: cannot satisfy -package-id mtl-2.2.1-b6b36d751642fde56f037818eb788105
    (use -v for more information)
Failed to install alex-3.1.7
Build log ( /Users/<User>/.cabal/logs/alex-3.1.7.log ):
[1 of 1] Compiling Main             ( /var/folders/_m/1h8dd_fd3dx3qvtq87_c4q480000gn/T/cabal-tmp-85241/alex-3.1.7/dist/setup/setup.hs, /var/folders/_m/1h8dd_fd3dx3qvtq87_c4q480000gn/T/cabal-tmp-85241/alex-3.1.7/dist/setup/Main.o )
Linking /var/folders/_m/1h8dd_fd3dx3qvtq87_c4q480000gn/T/cabal-tmp-85241/alex-3.1.7/dist/setup/setup ...
Configuring alex-3.1.7...
Building alex-3.1.7...
Preprocessing executable 'alex' for alex-3.1.7...
<command line>: cannot satisfy -package-id QuickCheck-2.8.1-8f3828a4d4d9dac78c5d3db525b2f0fe
    (use -v for more information)
cabal: Error: some packages failed to install:
alex-3.1.7 failed during the building phase. The exception was:
ExitFailure 1
happy-1.19.5 failed during the building phase. The exception was:
ExitFailure 1

Reimplements Dockerfile with the following changes:

• Uses the haskell Docker image as a base image, so we don't have to worry about installing Stack
• Uses multistage builds, to avoid bulking the final image with compile-time artifacts.
  • Currently, the gibiansky/ihaskell:202108170429266e69c2 image is 6.26 GB on my computer, after a docker pull
  • This PR builds a docker image that's 1.93 GB. Note: the haskell:8.10.4 base image we're using is 1.52 GB, so this PR builds a docker image that only adds 400 MB to the base image!
  • This should greatly improve download times
• Writes ~/.stack/global-project/stack.yaml to have the same resolver as the in-repo stack.yaml file (resolves https://github.com/gibiansky/IHaskell/issues/1251#issuecomment-899985184)
• Adds a --env-file flag, to fix using stack from the haskell base image (the stack exec env command called here was reinstalling GHC; didn't seem to be an easy way to tell it to use --system-ghc)

TODO:

• [ ] Fix syntax highlighting
  • Syntax highlighting is no longer working. I'm looking at the old version of the Dockerfile, and I'm not sure how it does syntax highlighting, since it's not COPYing the jupyterlab-ihaskell directory. @vaibhavsagar @gibiansky @gregwebs do you know how the current Docker image gets syntax highlighting? (Possibly related: https://github.com/gibiansky/IHaskell/issues/1199)

(ihaskell-venv) [[email protected] zeromq-4.0.5]# PKG_CONFIG_PATH=/e2e/lib/zeromq4/lib/pkgconfig:/e2e/lib/zeromq4/lib/pkgconfig:/usr/share/licenses/pkgconfig:/usr/share/pkgconfig:/usr/share/doc/pkgconfig cabal install ihaskell --extra-lib-dirs=/e2e/lib/zeromq4/lib --extra-include-dirs=/e2e/lib/zeromq4/include

...

Loading package MissingH-1.3.0.1 ... linking ... done. [16 of 18] Compiling IHaskell.Eval.Evaluate ( src/IHaskell/Eval/Evaluate.hs, dist/build/IHaskell/Eval/Evaluate.o )

src/IHaskell/Eval/Evaluate.hs:404:16: Couldn't match type FilePath' with[Char]' Expected type: GHC.IO.FilePath Actual type: FilePath In the first argument of writeFile', namely(fpFromString $ directory ++ filename)' In a stmt of a 'do' block: writeFile (fpFromString $ directory ++ filename) contents In the second argument of ($)', namelydo { createDirectoryIfMissing True directory; writeFile (fpFromString $ directory ++ filename) contents }'

src/IHaskell/Eval/Evaluate.hs:566:40: Couldn't match type FilePath' with[Char]' Expected type: GHC.IO.FilePath Actual type: FilePath In the return type of a call of fpFromString' In the second argument of($)', namely `fpFromString filename' In a stmt of a 'do' block: contents <- readFile $ fpFromString filename Failed to install ihaskell-0.6.0.0 cabal: Error: some packages failed to install: ihaskell-0.6.0.0 failed during the building phase. The exception was: ExitFailure 1

This is a while out, but it might be nice to package IHaskell nicely for end-users that aren't as experienced with Haskell.

Other ideas:

• Make a Docker build (https://github.com/gibiansky/IHaskell/issues/136) so that it is easy to install on any Linux distribution
• Package as Mac app (that can hang around in the task bar and have it's own window instead of using a browser, by having our own webview or something.)
• Package for windows like Sage does (see here)
• Package for package managers?
  • Homebrew?
  • Debian's apt-get?

Not sure which of these are worthwhile, but might be a good effort to get IHaskell and Haskell out to a wider audience.

The $25 bounty on this issue has been claimed at Bountysource.

This commit removed -dynamic as a default flag https://github.com/IHaskell/IHaskell/commit/2891bc8498ce12ae2677d72c7252e2623158e378

The results of this change break IHaskell in two ways. First, it now links every package available statically on startup. This is fine for small projects. In a large project though, with 200 packages, I gave up after 15 minutes. Second, many of us use IHaskell with packages that rely on shared libraries. This stopped working entirely.

There's a related long discussion here https://github.com/IHaskell/IHaskell/pull/1252/files#diff-30f4411ae6041ebe629c055ae5c4407e47989f719a5733416bfa23454270ba65R130

The suggested solution is that users should add dynamic in this case. But there is simply no way for users to figure this out.

I had to run strace and notice among its 100Mb of output that ihaskell is stuck loading one object file after another and look at gdb to confirm that it's stuck in the linker pinning the CPU to 100%. No one is going to do this and then realize that -dynamic is needed in IHaskell.

But it's even worse for a small project. You will inexplicably get errors about shared libraries. And they'll be lazy! So most of the code works, until you trip over some code that uses the shared library and then you get an incredibly cryptic error about symbols missing. Users have absolutely no way of tracking this back to IHaskell, never mind realizing that they need to add -dynamic.

This was done to make building docker images simpler. https://github.com/IHaskell/IHaskell/pull/1252

All of this is a horrible cost to pay for some convenience when building docker images. It breaks the common case for a slightly more convenient corner case. The downside of this pull request is it breaks building the Docker images, would you mind fixing that @brandon-leapyear? You seemed to have a version that copies over the right files.

I suggest we add -dynamic back in. @junjihashimoto suggested the same in the discussion above.

I wonder is there any specific reason all those ihaskell-* packages are not submitted to Stackage?

In my workflow, I build and install a version of IHaskell locally, and also have multiple projects which need IHaskell. Each project has different dependencies, but they all have to be careful to use exactly the same versions for ihaskell-* packages (by including a rather long section in stack.yaml). It's difficult to maintain, and when I want IHaskell on a new machine, I also have to be careful to build and install the exact same version all my projects use. If all these packages' versions are locked in a Stackage snapshot, it will be far easier to simply use the same Stackage LTS version for all my projects.

When starting "ihaskell install --stack", the terminal gives an error "ihaskell: command not found". Please tell me how it can be fixed?

I got the following error when I started a kernel

[I 00:16:53.865 NotebookApp] Creating new notebook in
[I 00:16:54.596 NotebookApp] Kernel started: 8818e04f-cf57-49f7-848b-23868661e6d8, name: haskell
[W 00:16:54.688 NotebookApp] 404 GET /static/components/codemirror/mode/ihaskell/ihaskell.js (127.0.0.1) 4.710000ms referer=http://localhost:8888/notebooks/Untitled.ipynb?kernel_name=haskell
[I 00:16:54.801 NotebookApp] Adapting from protocol version 5.0 (kernel 8818e04f-cf57-49f7-848b-23868661e6d8) to 5.3 (client).
ihaskell: Not in scope: ‘IHaskell.Eval.Widgets.relayWidgetMessages’
No module named ‘IHaskell.Eval.Widgets’ is imported.

[I 00:16:57.587 NotebookApp] KernelRestarter: restarting kernel (1/5), new random ports
WARNING:root:kernel 8818e04f-cf57-49f7-848b-23868661e6d8 restarted
[I 00:17:04.631 NotebookApp] Saving file at /Untitled.ipynb

but after it automatically restarted, the error disappeared and the kernel ran without problems any more. so basically I still can use ihaskell, just the error is weird and a bit annoying. anyway it would be very nice if you could have a look.

versions:

• ihaskell: master branch
• ghcup: v0.1.18.0
• stack: 2.9.1
• python: 3.11.0
• system: wsl ubuntu jammy and I installed ihaskell following the readme

https://github.com/IHaskell/IHaskell/blob/7d0b9b070aa821db1a4d38826e146fd2c41d1c0b/flake.nix#L95-L111

I would like the Nix Flake to provide a parameterized derivation for a Jupyter environment, where the parameter is a list of the display packages that should be installed.

It seems like there is not yet a first-class way to do that in Nix, but the discussion is ongoing here. https://github.com/NixOS/nix/issues/4744