v2.3.0

Cuba v2.3.0

Diff since v2.2.0

Closed issues:

• Combining Cuba.jl with GPU (#36)
• Controlling Size of X Matrix using nvec (#37)
• change license to MIT? (#40)

Merged pull requests:

• Test on Alpine Linux (#35) (@giordano)
• Upgrade Documenter to v0.27 (#38) (@giordano)
• add support for userdata and the corresponding tests (#39) (@kunyuan)
• [CI] Move Aarch64 to Cirrus and add FreeBSD as well (#41) (@giordano)
• [CI] Remove Drone (#42) (@giordano)
• Switch license to MIT (#43) (@giordano)

v2.2.0

Cuba v2.2.0

Diff since v2.1.0

Closed issues:

• Cannot differentiate through cuhre. Explicit type inference (#28)
• Run benchmark codes (#30)

Merged pull requests:

• Switch CI to GitHub Actions (#31) (@giordano)
• More modular setup for Drone CI (#32) (@giordano)
• Update to Cuba v4.2.1 (#33) (@giordano)

v2.1.0

Cuba v2.1.0

Diff since v2.0.0

Closed issues:

• Integration on general domains (#20)

Merged pull requests:

• Add a way to manually build Cuba (#22) (@giordano)
• Add Drone CI jsonnet script and generate .drone.yml from it (#23) (@giordano)
• Install TagBot as a GitHub Action (#25) (@JuliaTagBot)
• Use JLL package to provide the library (#26) (@giordano)

v2.0.0

Breaking Changes

• Support for Julia 0.7 was dropped.

New Features

• When using the package in the REPL, the result of the integration now has a more informative description about the error flag.

v1.0.0

Breaking Changes

• Support for Julia 0.6 was dropped.
• Keyword arguments reltol and abstol are now called rtol and atol, respectively, to match keywords in isapprox function.
• Deprecated functions llvegas, llsuave, lldivonne, and llcuhre have been removed.

New Features

• The build script has been updated, now the package supports Linux-musl and FreeBSD systems.

v0.5.0

New Features

• The package now uses BinaryProvider.jl to install a pre-built version of the Cuba library on all platforms.

Breaking Changes

• Support for Julia 0.5 was dropped
• The default value of argument ndim has been changed to 2 in divonne and cuhre. These algorithms require the number of dimensions to be at least 2. Now setting ndim to 1 throws an error. Your code will not be affected by this change if you did not explicitely set ndim. See issue #14.