Version 1.1.6 - Bug Fix for Python 3.8 compatibility

Two new updates for this release!

1. GPI is now being packaged as a "noarch" package on conda-forge. This means that you'll be able to install GPI for any system in any environment with Python 3. However, native Windows functionality is still a work in progress — we are hoping to take care of this fairly soon in GPI 1.2, but for now, you can make use of the Windows Subsystem Linux installation method described in this post.

2. While we have already begun packaging GPI as noarch, there was a bugs present when trying to run with Python 3.8 that caused GPI_PROCESS type nodes to fail. This has been addressed in version 1.1.6, so things should behave normally.

As always, please let us know if you encounter any issues! We continue to work on more rigorous testing but for now expect that we may not have caught everything prior to a release.

Version 1.1.5 — Minor bug fix for drag/drop file loading

This release provides a minor bug fix — drag/drop of files to the canvas should now (as in v1.0 and before) lead to creation of the appropriate node for loading the file.

As before, you can modify these bindings in your .gpirc file by uncommenting lines in the [ASSOCIATIONS] section. This is useful if you want to use nodes not contained in the core nodes library to load a certain file type. By default these lines are commented out and do nothing, but you may still wish to bring them in line with this bug fix by changing them to point to gpi_core rather than just core. In other words,

#BIND_7 = ('.npy', 'core.fileIO.ReadNPY', 'File Browser')

should become

#BIND_7 = ('.npy', 'gpi_core.fileIO.ReadNPY', 'File Browser')

You can do this manually in .gpirc or back it up, generate a new .gpirc with GPI v1.1.5, and then copy over any settings from your previous file.

GPI Version 1.1

This release should feel much the same as GPI v1.0, but will look a bit different thanks to support for qt5. Although we've attempted to preserve backwards compatibility with qt4, we are no longer supporting it and recommend upgrading.

In addition, we've added support for packaging node libraries via conda-forge to accompany GPI's move to conda-forge. The core node library has also moved to conda-forge in its latest release, v2.0. GPI users are still welcome to keep their libraries locally, as before. If you're interested in adding your node library to conda-forge, please contact us.

This release also features minor changes to the gpi_make script:

• Addition of a new flag, --ignore-system-libs, that avoids linking to libraries outside the conda environment. This can be called during a gpi_make and can be used in addition to --ignore-gpirc.
• Updated text output when running gpi_make to better describe the build paths being specified.

v1.0.6rc

I've tweaked the build scripts to try to make things more robust. For now I've only pushed new binaries to my personal branch (aganders3) on Anaconda.org. MacOS is available now, and I will build equivalent Linux versions today.

I made builds for Python 3.4-3.6 as it seems there is no Python 3.7 build for pyqt4. Once we merge the qtpy patch we can relax the pyqt dependency and enable support for Python 3.7 at the same time.

Regarding NumPy versions, I am following advice I got from the Anaconda mailing list and building with v1.11, which should be backward-compatible with all current NumPy versions from what I can tell (the issues we've seen apparently come from linking a newer version and trying to run with an older version).

If anyone wants to test can you try to install and let me know if anything is broken or if any dependencies are missing?

I'm curious how it works using GPI_Install_Latest.sh (see the attached version or modify yours to pull from my channel) or the following command in a new conda env.

conda install gpi gpi-core-nodes -c aganders3 -c conda-forge -c defaults

If there are no problems I'd like to push these binaries (framework and nodes) to the main channel and possibly archive the older builds (give them a different label) that contain some issues.

ISMRM 2019 Demo

This release contains all example material necessary for the ISMRM 2019 Demo session. The packages are meant for quick deployment on an Amazon AMI, Virtual Box or VMWare environment.

Notes:

• the OVA package contains pre-installed vm tools for VirtualBox and VMWare
  • Xubuntu 18.04 Desktop
  • 40GB hdd, 2GB memory, and 2-CPUs
  • use the startGPI script on the desktop to launch GPI
• Amazon AMI
  • ami-02203acf16dd4a983 (US East, Ohio)
  • ami-0080216529b04243b (US West, Oregon)
  • min requirements for demo: 1GB memory, 30GB ssd (e.g. all free tier EC2 options)
  • VNC is auto-started, to connect use the following formatted command:
    • ssh -i <your ec2 instance pem file> -L 5900:localhost:5900 ubuntu@<instance domain>
    • for example:
    • ssh -i ~/GPILAB_DEMO.pem -L 5900:localhost:5900 [email protected]
  • use the startGPI script on the desktop to launch GPI

GPI Framework v1.0.3

It's been a while since the last release, but here we go!

New features are minimal, but there are significant changes to the documentation. Check out the all-new GPI Framework Developer's Guide. Now you too can make improvements to the GPI Framework.

Of course, there are also some bug fixes! Enjoy!

GPI packaging is in a bit of a transitional state. Unfortunately we no longer have the personnel to maintain the macOS App and VM distributions at this time. The best way to get the latest version is to use the GPI Install Script, or run conda update gpi -c gpi in your Anaconda GPI environment.

Updated Script Installer and VM

Updates:

• Pulls the latest GPI Anaconda package
  • Circumvents a current issue with qt/libpng package dependencies in their latest builds (see #3).

MD5 (GPI_Install_Latest.sh) = 3c34066a045f954ba0e1d27477fa23bb
MD5 (GPI_Stack_1.0.2.vmwarevm.zip) = 8888b51f85407cc9d3e6f93d7ffa8638

commit 1a9f197121e6d5bfc2b9393bcfc57c5b82c8f5a7

GPI v1.0.0-rc Installer Packages

MD5 (GPI_1.0.0-rc.dmg) = 1ca1b7f2e779c06810a302c91ad44f28
MD5 (GPI_Install_Latest-rc.sh) = b1c5ed182324b715c3a83c39b9fbb05a
MD5 (GPI_VM_1.0.0-rc.zip) = d664cb72fb7a0efe762da1475049b9c8

GPI v0.5 Installer Packages

MD5SUMs:
3eb2574e9abdc35a51fa0e74657fbaff GPI_Stack_0.5.0-n1.pkg
a495567aa27e760df07cb65c56e051da GPI_Stack_0.5.0-n1.run
13040ae7519f98915b48b4c6a403612f GPI_Stack_0.5.0.vmwarevm.zip