The midyear release of Anaconda Distribution is available for download. The Anaconda 2019.07 release focuses on Conda performance. We’ve heard your frustration about lengthy install times for packages when using Conda, and the Conda 4.7 release aimed to address those concerns. The short version is that Conda install times have decreased by more than half for large packages such as PyTorch. The details are in the “How We Made Conda Faster in 4.7” blog post. Anaconda Distribution and Miniconda include Conda 4.7.10 which improves environment management and error messages.
In addition to Conda 4.7, we released Conda build 3.18 to work with the new .conda file format. This file format uses the zstd compression library, leading to smaller downloads and faster decompression. Finally, another major change to Conda is a shift to a quarterly release cycle – previously we released conda based on the features added. With a regular release cycle, we will provide smaller (feature-wise) but more frequent releases.
Conda constructor, the cross platform installer tool, now supports building installers with the new .conda file format as well as noarch packages. Anaconda 2019.07 is the first installer release to include noarch packages which helps reduce our overall maintenance burden and helps us scale up our build efforts. This makes it more feasible for users to create installers from conda-forge packages, where noarch is especially encouraged for efficiency with build resources. The new file format reduced the size of Windows and Linux installers significantly – for Anaconda3 on linux-64, this was a drop from 654 MB to 517 MB. Because the MacOS .pkg GUI installers must unpack each archive before creation of the installer, they do not benefit from the new file format’s better compression.
At AnacondaCON in April, we announced our partnership with JetBrains to deliver Anaconda Distribution integration with PyCharm. The Docs team has updated the documentation with PyCharm instructions as well as adding new tutorials for Conda build. The Docs team has been reorganizing and updating content and improving navigation over several releases. We hope that these improvements make them an even more useful resource.
As with every release, we’ve updated packages in the Anaconda Distribution installer. We continue to update packages throughout the year and make them available at repo.anaconda.com. Package updates include R packages, ML/AI packages such as TensorFlow and Keras, Dask and many more. You can check on the latest package updates with the RSS feed.