Usage in R via reticulate

message_ix and ixmp are fully usable in R via reticulate, a package that allows nearly seamless access to the Python API. See Installation for instructions on installing R and reticulate. No additional R packages are needed. [1]

Once installed, use reticulate to import the Python packages:

library(reticulate)
ixmp <- import("ixmp")
message_ix <- import("message_ix")

This creates two global variables, ixmp and message_ix, that can be used much like the Python modules:

mp <- ixmp$Platform(name = "default")
scen <- message_ix$Scenario(mp, "model name", "scenario name")
# etc.

See the R versions of the Austria tutorials for full examples of building models.

Some tips:

  • If using Anaconda, you may need to direct reticulate to use the Python executable from the same conda environment where message_ix and ixmp are installed. See the reticulate documentation for usage of commands like:

    # Specify python binaries and environment under which messageix is installed
use_condaenv("message_env")
# or
use_python("C:/.../anaconda3/envs/message_env/")

  • As shown above, R uses the $ character instead of . to access methods and properties of objects. Where Python code examples show, for instance, scen.add_par(...), R code should instead use scen$add_par(...).

  • MESSAGEix model parameters with dimensions indexed by the year set (e.g. dimensions named year_act or year_vtg) must be indexed by integers; but R treats numeric literals as floating point values. Therefore, instead of:

    ya1 = 2010
ya2 = c(2020, 2030, 2040)
ya3 = seq(2050, 2100, 10)

# ...store parameter data using year_act = ya1, ya2, or ya3

    …use as.integer() to convert:

    ya1 = as.integer(2010)
ya2 = sapply(c(2020, 2030, 2040), as.integer)
ya3 = as.integer(seq(2050, 2100, 10))