MAgPIE 4.0 - data download error

[jansteinhauser]

Hello again!
I want to do a few test runs on v4.0 to compare them to the runs released with the 4.0 paper in 2019. However, when I try downloading the required inputs, the script stops at f12_interest_rate_bound.cs3 because:

Error in write.magpie(round(read.magpie(input_file), round), output_file) : Input data seems to be sparse but cs3 does not support sparse data. Please use cs4 instead!

Any ideas what I can do about this? Could this be because some R packages are too new for this script?

Thanks and have a nice weekend!
Jan

[abhimishr]

I have a dirty fix for you, maybe @tscheypidi can suggest an elegant way of doing it because this concerns backward compatibility as well as reproducibility to a small extent.

1. Replace your download_data_only.R script with

require(lucode2)
require(gms)

source("scripts/start_functions.R")
source("config/default.cfg")
download_and_update(cfg)

2. Replace start_functions.R script with one of the most recent one (maybe from the master branch)

Execute download data step - will throw some errors but the model still compiles, maybe give it a try (hoping the run goes through).

Cheers,
Abhi

[jansteinhauser]

Hey, thanks for this. As a test, instead of step 1, I actually just added require(gms) to the download_data_only.R script and even that (plus step 2) was enough. Any specific reason for the other changes to the script?

[acsoterroni]

Hello,

I'm a new MagPIE user! To run version 4.0 I had to:

1. Replace lucode by lucode2 everywhere (I used 'perl' command);
2. Add require(gms) to ./script/start/download_data_only.R script;
3. Replace ./script/start_functions.R script by the one from version 4.3.5.

I was able to generate the fulldata.gdx but my default_validation.pdf has only 3 pages! However, I was able to generate a complete PDF file for the version 4.3.5. I'm running MagPIE on Unix.

Best,
Aline

[abhimishr]

Welcome to MAgPIE @acsoterroni!

Unfortunately, generation of validation PDF is bit more tricky to get right on local machines. See discussion from #333 (reply in thread) onwards. The log file in your model output folder might provide an error message on what went wrong. Please don't hesitate to open a new discussion thread about this, and then we can specifically have a look at it.

[abhimishr]

@jansteinhauser I just loaded up some library calls from a personal start script hence the calls to both gms and lucode2. Probably just calling gms is enough as you said!

cheers,
Abhi