~10x speedup from lazy-loading standard_vars

gistfile1.txt

library(microbenchmark)

# PEcAn.utils::to_ncvar, revision 79ef207
to_ncvar_current <- function(varname,dims){
  standard_vars <- read.csv(system.file("data/standard_vars.csv", package="PEcAn.utils"), stringsAsFactors = FALSE)
  var <- standard_vars[which(standard_vars$Variable.Name == varname),]
  #check var exists
  if(nrow(var)==0){
    PEcAn.logger::logger.severe(paste("Variable",varname,"not in standard_vars"))
  }
  
  dimset <- var[,c("dim1","dim2","dim3","dim4")]
  dim <- dims[which(names(dims) %in% dimset)] #subset list of all dims for this variable
  #check that dim isn't 0
  if(length(dim)==0 || is.null(dim)){
    PEcAn.logger::logger.severe(paste("No dimensions were loaded for",varname))
  }
  
  units = as.character(var$Units) #if the units are a factor the function fails
  longname <- as.character(var$Long.name)
  
  ncvar <- ncdf4::ncvar_def(name = varname, units = units, longname = longname, dim = dim, -999, prec = "double") 
  
  return(ncvar)
}

# PEcAn.LZutils::to_ncvar
# Package modifications to lazy-load rather than reread standard_vars every time
to_ncvar_lazy <- function(varname,dims){
  var <- PEcAn.LZutils::standard_vars[which(PEcAn.LZutils::standard_vars$Variable.Name == varname),]
  #check var exists
  if(nrow(var)==0){
    PEcAn.logger::logger.severe(paste("Variable",varname,"not in standard_vars"))
  }
  
  dimset <- var[,c("dim1","dim2","dim3","dim4")]
  dim <- dims[which(names(dims) %in% dimset)] #subset list of all dims for this variable
  #check that dim isn't 0
  if(length(dim)==0 || is.null(dim)){
    PEcAn.logger::logger.severe(paste("No dimensions were loaded for",varname))
  }
  
  units = as.character(var$Units) #if the units are a factor the function fails
  longname <- as.character(var$Long.name)
  
  ncvar <- ncdf4::ncvar_def(name = varname, units = units, longname = longname, dim = dim, -999, prec = "double") 
  
  return(ncvar)
}

# Generate some fake data to write
 dims <- list(
	lon = ncdf4::ncdim_def(name = "lon", vals = -88.1, units = "degrees_east") ,
	lat = ncdf4::ncdim_def(name = "lat", vals = 40.08, units = "degrees_north"),
	time = ncdf4::ncdim_def(name = "time", vals = 3, units = "days since 1700-01-01 00:00:00 UTC"))

microbenchmark(current = to_ncvar_current("NPP", dims), lazy = to_ncvar_lazy("NPP", dims))

Unit: microseconds
    expr      min        lq      mean    median        uq      max neval
 current 2675.269 2726.2315 3509.4568 2776.9280 2839.1235 69665.08   100
    lazy  208.281  221.5785  480.6783  227.5105  241.9015 21925.81   100