Skip to content

Latest commit

 

History

History
459 lines (371 loc) · 18.5 KB

README.md

File metadata and controls

459 lines (371 loc) · 18.5 KB
Raw

rwrfhydro

Travis-CI Build Status

A community-contributed tool box for managing, analyzing, and visualizing WRF Hydro (and HydroDART) input and output files in R.

Purpose

Intentionally, “rwrfhydro” can be read as “our wrf hydro”. The purpose of this R package is to focus community development of tools for working with and analyzing data related to the WRF Hydro model. These tools are both free and open-source, just like R, which should help make them accessible and popular. For users new to R, several introductory resources are listed below.

The purposes of this README are 1) to get you started using rwrfhydro and 2) to explain the basics (and then some) of how we develop the package so you can quickly start adding your contributions.

Table of Contents

Installing

Installing rwrfhydro (not on CRAN) is facilitated by the devtools package (on CRAN), so devtools is installed first. The following is done for the initial install or to update the rwrfhydro package.

install.packages("devtools")
devtools::install_github("NCAR/rwrfhydro")

The very first time this is run, it can take a while to install all the package dependencies listed as “Imports” in the DESCRIPTION file.

To check for updates once rwrfhydro is loaded run CheckForUpdates().

To install other branches than master and perhaps from your own fork:

devtools::install_github("username/rwrfhydro", ref='myBranch')

Importantly, beta functionality can be installed using:

devtools::install_github("NCAR/rwrfhydro", ref='devBranch')

We are finally gaining some Windows users and have attempted to improve portability of rwrfhydro to this system. A primary dependence of rwrfhydro is ncdf4. This ncdf4 package binary can be installed in the following way: First, obtain the binary from http://cirrus.ucsd.edu/~pierce/ncdf/. Then in an R session

install.packages(file.choose(), repos=NULL, type = "binary")

R will open a window for you to choose the downloaded zip file and will install it.

Using

After the one-time install or any subsequent re-installs/updates, simply load the rwrfhydro library in an R session:

library(rwrfhydro)

and now the package (namespace) is available.

Online vignettes (or in R browseVignettes("rwrfhydro")) are probably the easiest way to get in-depth, thematic overviews of rwrfhydro functionality.

To get package metadata and a listing of functions: library(help=rwrfhydro). Just the functions: ls('package:rwrfhydro'). For specific functionality see function help (e.g. ?VisualizeDomain or help(VisualizeDomain)).

Developing and bug reports

Bugs are to be reported here. If you want to help solve bugs and fixes into the code, please continue reading about developing.

There are four main aspects of developing the code base and contributing:

  • Version control for collaboration: Not terribly interesting but incredibly useful. For those new to the Git/Github process, it can be a bit daunting. Please contact us for some help, we do want to get your useful code into the main repository!

  • R Packaging: Again, not very interesting, but critical for creating the extrememly useful nature of R packages. Fortunately, the devtools package simplifies packaging tremendously and so figures prominently in the development process we sketch below. The main details have been sorted out, contributing new functions is generally fairly easy.

  • Our best practices: This ranges from fundamental to fussy.

  • Getting in touch with the community: Lots of aspects of this tool are under active development. Dont duplicate efforts, extend them! We will establish a site for communicating development tasks and statuses on the github wiki.

R packageing and some version controll are treated by Hadley Wickham’s book on R Packages. Specific sections of this book are linked below. Further resources on R package development are listed at the end of this page.

Version control for collaboration: Github

Instead of going straight to developing, we recommend that you install rwrfhydro using devtools::install_github('NCAR/rwrfhydro') first, because this streamlines the installation of package dependencies. Note that devtools::install_github('NCAR/rwrfhydro') installs rwrfhydro into you default library path (.libPaths()[1]) and that the source code is not included.

The very best way to obtain and edit the source is to “fork” rwrfhydro on github and then clone your the repository to your machine(s) of choice. You edit your fork and, when it’s ready, you submit a pull request to get the changes back to the main (upstream) fork of rwrfhydro. More details are provided below. Your cloned git repository is not in your default R library path (.libPaths()[1]), but somewhere else where you choose to keep your development code. However, devtools allows you to build your development package into your library path. This means that after you add some code locally, you can library(rwrfhydro) from other R sessions on that machine with your changes appearing in the package. The basic use of devtools is outlined below. It greatly stream lines all aspects of developing R packages and is highly recommended. Particularly, it make is easy to go from github or local changes to an R package.

Forking and cloning

Please fork the repository to contribute. A fork is a separate copy of the main repository on which you have write permissions. Note that you do not have write permissions on any other fork of the repository. Forking is trivial in Github. You have to have a free (for open-source repositories) account to fork on github.

Next you’ll clone your fork to your local computer and you’ll to interact between your forked repository on github, which is called “origin”. The repo “NCAR/rwrfhydro” is known as the “upstream” fork. This is the “official” repo. It’s also called upstream because changes to it should always flow to all other repos so that they can easily sync their separate changes back to it. Keep your fork sync’d with upstream as much/often as possible to avoid painful merges, github notifies downstream forks of changes to upstream.

Because you dont have write permissions to “upstream” (or any other fork), you have to request that your changes be pulled upstream. This is done via a pull request on github (website). We give some tips below and give a general overview of forking on github in this document.

devBranch and pull requests

We maintain two main branches or rwrfhydro: master and devBranch. You should never work on the master branch. All changes have to pass through the devBranch before going in to the master and this is controlled by the package maintainer. Therefore, devBranch is where your pull requests will go. Other barnches on your fork are up to you. How you get your code into your fork’s devBranch is your choice. One suggestion is to work on your personal branch. Then when various files are ready to be contributed to devBranch, you first do git checkout devBranch then followed by git checkout myBranch -- path/to/file for each file you desire to copy from myBranch into devBranch. Finally the git add and git commit formally put these files into devBranch. Some more details on using git are provided in the workflow overview below.

Not using Github.

This is not recommended, but might be possible. It will certainly hinder you interaction with the upstream repo.

Travis-CI and R CMD check

The rwrfhydro repo is configured to build on a third-party virtual linux machine with every push or pull request to the master or devBranch branches. This service is known as Travis-CI (continuous integration). This means your pull requests are automatically checked by R CMD check, this keeps errors from creeping into the upstream code. There are a variety of hurdles to getting code to build on Travis-CI, including installing requisite system and R packages, which can be challenging but worth it for the debugging provided by automated builds in conjunction with R CMD check.

R CMD check accepts a variety of arguments. Ultimately, it 1) checks the source for consistency including across platforms (Windows, OSX, linux), 2) runs all specified code tests, essentially regression tests, 3) runs all the examples provided in the documentation, and 4) builds all the vignettes. Currently, we are skipping vignette building until we can streamline several of these.

You can configure your own fork to build on Travis-CI and you can push frequently to check for errors. This is nearly identical to (slightly more stringent than) running devtools::check(), but all you have to do is push your commits.

Workflow: git, R Packaging, and you

Workflow is approximately this:

  • Fork project on github

  • Clone your github fork to your local machine(s)

  • Set the upstream repo (git remote add upstream https://github.com/NCAR/rwrfhydro - different syntax for ssh access)

  • Checkout your devBranch (git checkout devBranch)

  • Development cycles:

    • Optional: Create a topic branch off of devBranch in git (e.g. git checkout -b myBranch) and push this to github (git push origin myBranch).
    • Write code (in these dirs: R/, NAMESPACE, src/, data/, inst/).
    • Write documentation (in these dirs: man/, vignettes/).
    • Write tests (in this dir: test/).
    • Document and check with devtools: devtools::document(); devtools::check_man(); devtools::check()
    • Commit to your branch with git. (git commit -am 'Some cool features were needed.')
    • Push to github (git push origin branch). If Travis-CI is configured, this can trigger an R CMD check on Travis.

  • To get code back to the main reposiory/fork:

    • If on myBranch: You probably want to mege devBranch into myBranch: git pull upstream devBranch.
    • If on myBranch: checkout devBranch. The either wholsale merge your work from origin (git pull origin myBranch) or cherry pick files (git check out myBranch -- path/to/file for eac file) it into devBranch.
    • if on myBranch: commit, git commit -am'Fixes and new features'.
    • If not previously on myBranch and did not do a wholesale merge in previous step: Sync with the “upstream” devBranch: See here. If upstream repo is set: git pull upstream devBranch.
    • Push your devBranch to your fork on github: git push origin devBranch.
    • Submit a pull request on github on devBranch to upstream (NCAR/rwrfhydro).

Our best practices

R package best practices and code style

http://r-pkgs.had.co.nz/r.html

Organizing functions

http://r-pkgs.had.co.nz/r.html#r-organising

  • Do NOT put all functions in a single file, nor each in their own file. Functions should be grouped by files and may occasionally need moved to new or different files as new functions are written.
  • File names end in .R and are all lowercase with _ used to separate words. (All lowercase (except the .R) helps ensure compatibility with Windows developers.)

R code style

http://r-pkgs.had.co.nz/r.html#style

  • Generally follow Google’s R style guide with preference for variableName (first-lower camel case) over variable.name (period distinction). Note that functions are first-upper camel case, e.g. FunctionName. https://google-styleguide.googlecode.com/svn/trunk/Rguide.xml
  • Variables are nouns. Functions are verbs.
  • Lots of other style considerations to learn: indents, braces, line length, assignment, comment style.

Packages are NOT scripts

http://r-pkgs.had.co.nz/r.html#r-differences

  • Don’t use library() or require(). Use the DESCRIPTION to specify your package’s requirements.
  • Use package::function() to use function from external packages. Make sure the package and version are listed in DESCRIPTION.
  • Never use source() to load code from a file. Rely on devtools::load_all() to automatically source all files in R/.
  • Don’t modify global options() or graphics par(). Put state changing operations in functions that the user can call when they want.
  • Don’t save files to disk with write(), write.csv(), or saveRDS(). Use data/ to cache important data files.

Documentation with roxygen

http://r-pkgs.had.co.nz/man.html

Once you get used to this, you will love writing documentation as you go for your R functions.

  • Roxygen comments start with #’ and come before a function. All the roxygen lines preceding a function are called ablock. Each line should be wrapped in the same way as your code, normally at 80 characters.
  • Blocks are broken up into tags, which look like @tagName details. The content of a tag extends from the end of the tag name to the start of the next tag (or the end of the block). Because @ has a special meaning in roxygen, you need to write @@ if you want to add a literal @ to the documentation (this is mostly important for email addresses and for accessing slots of S4 objects).
  • Each block includes some text before the first tag. This is called the introduction, and is parsed specially:
  • The first sentence becomes the title of the documentation. That’s what you see when you look at help(package = mypackage) and is shown at the top of each help file. It should fit on one line, be written in sentence case, and end in a full stop.
  • The second paragraph is the description: this comes first in the documentation and should briefly describe what the function does.
  • The third and subsequent paragraphs go into the details: this is a (often long) section that is shown after the argument description and should go into detail about how the function works.
  • All objects must have a title and description. Details are optional.
  • GetPkgMeta: NOTE that this function only works for packages not installed by devtools::load_all(). The function analzes the @keywords and @concepts tags supplied in the roxygen documentation to categorize and relate functions and objects in rwrfhydro (or any other package). Keywords follow R conventions (see this obscure link which I need to find). Concepts are customized to rwrfhydro. The current list of keywords and concepts can be gathered from the latest version of devBranch, using the following command, which gives as of this writing:
> GetPkgMeta(listMetaOnly=TRUE)
-----------------------------------
rwrfhydro concepts
-----------------------------------
Ameriflux
DART
data
dataAnalysis
dataGet
dataMgmt
geospatial
getData
GHCN
modelEval
MODIS
ncdf
nudging
plot
SNODAS
SNOTEL
Streamflow
usgs
usgsStreamObs

-----------------------------------
rwrfhydro keywords
-----------------------------------
data
database
hplot
internal
IO
manip
smooth
ts
univar
utilities

Objects in rwrfhydro

We will probably need to develop some s3 classes or reuse some from other packages. List of possible objects: gaugePts object for organizing “frxst points”, both locations and data.

Graphics

We need to resolve if we are going to use base graphics or ggplot or both. I’m leaning towards both. Not all plotting routines have to always be available for a given function, but I think that both will probably develop over time.

Because ggplot has a big learning curve, we can return closures which 1) provide tweakability for basic things to be adjusted in the plot make the plot when called, 2) which return the basic ggplot object which can then also be extended with ggplot commands. I made an example of this in VisualizeDomain.R for ggmap/ggplot objects.

R Package development resources

Introductory R resources (somewhat random)