Advice

I have command line arguments or configuration parameter stored as strings. I want to use them as numbers, dates, boolean etc. A conversion functionality, using C is usually provided by the sscanf(...) function. The function realizes type conversion and syntax check with a hint to the error location in the incoming string, using format strings to define the conversion context. I know there are POSIX functions to scan strings for date and time formats available (strptime(...)) and the common conversions as.numeric(...)

Is there a similar function like sscanf(...) available for R?

--- @margusl -- use case request -----------------------

In the INI file below, the important lines are:

[PROCESS]
FEATURE.HDR = ACTIVE:NORM:SAMPLES:FILE
FEATURE.FMT = %s:%d:%d:%s"

The pseudo function cfg.get.table(...)

# Psedo functionality
scan <- sscanf(format, str)

# with 
sscan <- function(format,str) {
  ...
  return(list(error=str OR NULL, data=list(...))  
}

sketches the expected functionality. I added the docopt part to explain the intended CLI/INI functionality for @G. Grothendieck. The question is about an existing sscanf(...) R feature. I am aware that it could be done using Rcpp, but I am not sure how to generate the list from format string the without writing a format parser for the '%XX' tags.

Docopt specification:

usage:
    tool (-h | --help)
    tool (-v | --version)
    tool \
         --config            <CONFIG.FILE> \
         --work-path         <WORK.PATH>   \
         --feature-set       <FEATURE.SET> 
    ...
options:
  -h, --help                        Show this screen.
  -v, --version                     Show version.
  --config <CONFIG>                 Configuration file.
  --work-path <WORK.PATH>           Work path, usually the project path
  --feature-set <FEATURE.SET>       Feature set chosen from config (input files)
...

Configuration file:

[COMMON]
...

[CLI.TO.CONFIG]

# CLI to CFG transformation rules
WORK.PATH    = /PATH/PROJECT
FEATURE.SET  = /PROCESS/FEATURE.SET
ALGORITHM    = /PROCESS/ALGORITHM
...

# PATH Settings addressing the project setup
[PATH]
PROJECT   = ${/SYSTEM/HOME}/project/data
INCOMING  = ${/PATH/PROJECT}/Incoming
...

# Parameter needed for the process
[PROCESS]
OVERWRITE         = OFF
USE.SUBSETS       = OFF
SAMPLE.SIZE       = 10000
ALGORITHM         = RPART
FEATURE.SET       = FEATURE.GRAY.MIN
FEATURE.HDR       = ACTIVE:NORM:SAMPLES:FILE
FEATURE.FMT       = %s:%d:%d:%s
...

[FEATURE.GRAY.MIN]
SPIX = ON:0:1000:${/PROCESS/SPIXEL.RASTER}
GRAY = OFF:1:2000:${/PATH/INCOMING}/ge-feature-gn.tif
VVIX = ON:0:1000:${/PATH/INCOMING}/ge-feature-b3n-vdni.tif
...

R-Pseudo Code

require(configr)
require(docopt)
source('lib/helper.R')
# Set the CLI specification and the configuration files
DOC.FILE <- 'etc/tool.docopt'
CFG.FILE <- 'etc/tool.ini'

test.args <- c(
  "--feature-set", "FEATURE.GRAY.MIN",
  "--config", CFG.FILE,
  "--work-path", getwd(),
  ...
) 

# Parse the command line parameter 
CLI <- cli.parse(DOC.FILE, test.args)
if (! is.null(CLI$error) ) {
    io.error(FALSE, "Invalid command line options!\n DETAILS: %s",
             CLI$error)
}

# Transfer CLI to CFG and resolve interpolated vars 
CFG <- cfg.parse(CFG.FILE, CLI, T)
if (! is.null(CFG$error) ) {
    io.error(FALSE, "Invalid CONFIGURATION: '%s'!\n DETAILS: %s",
             CFG.FILE, CFG$error)
}

# FT.SET addresses [PROCESS] 
#.   -> FEATURE.SET = FEATURE.GRAY.MIN'
FT.SET <- cfg.get.parameter(CFG, '%s', '/PROCESS/FEATURE.SET')
# FT.HDR addresses [PROCESS] 
#    -> FEATURE.HEADER = 'ACTIVE:METHOD:SAMPLES:FILE'
FT.HDR <- cfg.get.parameter(CFG, '%s', '/PROCESS/FEATURE.HDR')
# FT.FMT addesses [PROCESS] 
#.   -> FEATURE.HEADER = '%s:%d:%d:%s'
FT.FMT <- cfg.get.parameter(CFG, '%s', '/PROCESS/FEATURE.FMT')

...
# FT.TAB addresses row entries [FEATURE.GRAY.MIN]
# NAME ON/OFF METHOD:INT SAMPLES:INT FILE:STRING 
FT.TAB <- cfg.get.table(CFG, FT.SET, FT.HDR, FT.FMT)
...

# Function could look like this:
# config - nested list containing the config ...INI
# section - section name addressed by the feature set 
# header - header of the expected data frame
# format - format string of the expected layout 
cfg.get.table <- function(config, section, header, format) {
    sec  <- config[[section]]
    rows <- names(sec)
    thdr <- unlist(strsplit(value,':'))
    df   <- data.frame()
    for(ix in 1:length(rows)) {
      id  <- rows[ix]
      str <- sec[[id]]
      # --------- EXPECTED FUNCTION ------------------  
      scan <- sscanf(format, str)
      # ----------------------------------------------
      if (! is.null(scan$error) {
          stop($scan$error)
      } 
      if (ix == 1) {
         df <- data.frame(id=id, unlist(scan$data))
      }  else {
         df <- rbind(df,data.frame(id=id, unlist(scan$data)))
      }
   } 
   # Generate a well defined header as defined in the config
   names(df) <- c('ID', thdr);
   return(df)
}
Improve this question