release 1.7.1 of PEcAn

[robkooper]

Fixes

• Fixed issue that prevented modellauncher from working properly Issue with modellauncher #2262
• Use explicit namespacing (package::function) throughout PEcAn.meta.analysis. Otherwise, many of these functions would fail when trying to run a meta-analysis outside of the PEcAn workflow (i.e. without having loaded the packages first) (PEcAn.DB and PEcAn.meta.analysis cleanup #2351).
• Standardize how PEcAn.DB tests create database connections, and make sure tests work with both the newer Postgres and older PostgreSQL drivers (PEcAn.DB and PEcAn.meta.analysis cleanup #2351).
• Meta-analysis = "AUTO" now correctly skips the meta analysis if the PFT definition has not changed (Meta-analysis AUTO broken ... again #1217).
• Replace deprecated rlang::UQ syntax with the recommended !!
• Explicitly use PEcAn.uncertainty::read.ensemble.output in PEcAn.utils::get.results. Otherwise, it would sometimes use the deprecated PEcAn.utils::read.ensemble.output version.
• History page would not pass the hostname parameter when showing a running workflow, this would result in the running page showing an error.

Changed

• Updated modules/rtm PROSPECT docs
• Updated models/sipnet/R/model2netcdf.SIPNET.R to address issues in PR Updates to improve time handling with SIPNET runs #2254
• Improved testing (Travis: Build against multiple R versions #2281). Automatic Travis CI builds of PEcAn on are now run using three versions of R in parallel. This should mean fewer issues with new releases and better backwards compatibility, but note that we still only guarantee full compatibility with the current release version of R. The tested versions are:
  • release, the current public release of R (currently R 3.5). Build failures in this version are fixed before merging the change that caused them. When we say PEcAn is fully tested and working, this is the build we mean.
  • devel, the newest available development build of R. We will fix issues with this version before the next major R release.
  • oldrel, the previous major release of R (currently R 3.4). We will fix issues with this version as time allows, but we do not guarantee that it will stay compatible.
• Reverting back from PR (Update to MAAT model wrappers #2137) to fix issues with MAAT wrappers.
• Moved docker files for models into model specific folder, for example Dockerfile for sipnet now is in models/sipnet/Dockerfile.
• PEcAn.utils:
  • Remove, or make "Suggests", a bunch of relatively underutilized R package dependencies.
• Add template for documentation issues and add button to edit book.
• Conditionally skip unit tests for downloading raw met data or querying the database when the required remote connection is not available.
• Reorganization of docker folder
  • All dockerfiles now live in their own folder
  • scripts/generate_dependencies.R is now used to generate dependencies for make and docker
• In PEcAn.DB::get.trait.data, if trait.names is NULL or missing, use the traits for which at least one prior is available among the input list of PFTs. (Previously, we were getting this from the PEcAn.utils::trait.dictionary, which we are trying to deprecate remove base/utils/data/trait.dictionary.csv #1747). (PEcAn.DB and PEcAn.meta.analysis cleanup #2351)
• Cleanup and improve logging and code readability in parts of PEcAn.DB related to getting trait data, including replacing many manual database queries with dplyr calls.
• Reorganization of PEcAn documentation in accordance with isue Documentation Outline #2253.
• SIPNET now is installed from the source code managed in git

Added

• Meta analysis functionality to not use greenhouse data.
• Dockerize the BioCro model.
• Added PRO4SAIL-D model, using existing 4SAIL src and coupling with PROSPECT-D Fortran code
• Models will not advertise themselvs, so no need to register them a-priori with the database advertise new models #2158
• Added simple Docker container to show all containers that are available (http://localhost:8000/monitor/). This will also take care of registering the models with the BETY database.
• Added unit tests for met2model.<MODEL> functions for most models.
• Added MAESPA model to docker build
• PEcAn has more robust support for RPostgres::Postgres backend. The backend is officially supported by db.query, and basic workflows run top-to-bottom with the Postgres backend. However, RPostgreSQL is still the default until we do more robust testing of all modules.
• PEcAn.DB::db.query now optionally supports prepared statements (check for sql injection paths; switch to using prepared sql statements (dplyr?) #395).
• New function PEcAn.DB::query_priors that expands the functionality of query.priors by (1) accepting PFTs by name or ID and (2) allowing the user to request all possible combinations of the input PFTs and traits (i.e. expand.grid(pfts, traits)) or just the pairwise combinations (i.e. pft[1]-trait[1], pft[2]-trait[2]). This function also comes with more robust error handling and a set of unit tests (PEcAn.DB and PEcAn.meta.analysis cleanup #2351).
• New function PEcAn.DB::query_pfts for finding PFT IDs and types from the PFT name and (optionally) model type (PEcAn.DB and PEcAn.meta.analysis cleanup #2351).
• Run Travis integration tests with both Postgres and PostgreSQL drivers (PEcAn.DB and PEcAn.meta.analysis cleanup #2351).
• New function PEcAn.utils::load_local reads Rdata files into a named list (instead of into the current environment).

Removed

• Removed unused function PEcAn.visualization::points2county, thus removing many indirect dependencies by no longer importing the earth package.
• Removed package PEcAn.data.mining from the Make build. It can still be installed directly from R if desired, but is skipped by default because it is in early development, does not yet export any functions, and creates a dependency on the (large, often annoying to install) ImageMagick library.
• Fully deprecate support for MySQL database driver. Now, only PostgreSQL (and, experimentally, RPostgres) are supported. With this, remove RMySQL dependency in several places.

Fixed

• Replace deprecated rlang::UQ syntax with the recommended !!
• Explicitly use PEcAn.uncertainty::read.ensemble.output in PEcAn.utils::get.results. Otherwise, it would sometimes use the deprecated PEcAn.utils::read.ensemble.output version.
• PEcAn.ED2::met2model.ED2 now skips processing of years for which all output files are already present (unless overwrite = TRUE). This prevents a lot of unnecessary work when extending an existing ED met record.

[dlebauer]

When posting the release notes,

• There are two sections 'fixed' and 'fixes' should these be merged?
• Also if it starts with new features (added) then changed then fixed, the exciting stuff comes first.