R/record_format.R

In naaccr: Read Cancer Records in the NAACCR Format

#' @noRd
naaccr_boolean12 <- function(x) naaccr_boolean(x, false_value = '1')


#' Wrapper to create "keep unknown" versions of functions
#' @noRd
keep_unknown <- function(fun) {
  formals(fun)[["keep_unknown"]] <- TRUE
  fun
}


#' `type`: name of the field type (see record_format)
#' `fun`: normal function used to parse the field (factors and sentinels are
#'   special).
#' `fun_unknown`: function used to parse the field when `keep_unknown = TRUE`
#'   for the reading/parsing function.
#' @noRd
type_converters <- rbindlist(list(
  list(type = "integer", fun = list(as.integer), fun_unknown = list(as.integer)),
  list(type = "numeric", fun = list(as.numeric), fun_unknown = list(as.numeric)),
  list(type = "character", fun = list(clean_text), fun_unknown = list(keep_unknown(clean_text))),
  list(type = "factor", fun = list(identity), fun_unknown = list(identity)),
  list(type = "sentineled_integer", fun = list(identity), fun_unknown = list(identity)),
  list(type = "sentineled_numeric", fun = list(identity), fun_unknown = list(identity)),
  list(type = "age", fun = list(clean_age), fun_unknown = list(keep_unknown(clean_age))),
  list(type = "icd_code", fun = list(clean_icd_code), fun_unknown = list(keep_unknown(clean_icd_code))),
  list(type = "postal", fun = list(clean_postal), fun_unknown = list(keep_unknown(clean_postal))),
  list(type = "city", fun = list(clean_address_city), fun_unknown = list(keep_unknown(clean_address_city))),
  list(type = "address", fun = list(clean_address_number_and_street), fun_unknown = list(keep_unknown(clean_address_number_and_street))),
  list(type = "facility", fun = list(clean_facility_id), fun_unknown = list(keep_unknown(clean_facility_id))),
  list(type = "census_block", fun = list(clean_census_block), fun_unknown = list(keep_unknown(clean_census_block))),
  list(type = "census_tract", fun = list(clean_census_tract), fun_unknown = list(keep_unknown(clean_census_tract))),
  list(type = "icd_9", fun = list(clean_icd_9_cm), fun_unknown = list(keep_unknown(clean_icd_9_cm))),
  list(type = "county", fun = list(clean_county_fips), fun_unknown = list(keep_unknown(clean_county_fips))),
  list(type = "physician", fun = list(clean_physician_id), fun_unknown = list(keep_unknown(clean_physician_id))),
  list(type = "override", fun = list(naaccr_override), fun_unknown = list(naaccr_override)),
  list(type = "boolean01", fun = list(naaccr_boolean), fun_unknown = list(naaccr_boolean)),
  list(type = "telephone", fun = list(clean_telephone), fun_unknown = list(keep_unknown(clean_telephone))),
  list(type = "count", fun = list(clean_count), fun_unknown = list(keep_unknown(clean_count))),
  list(type = "ssn", fun = list(clean_ssn), fun_unknown = list(keep_unknown(clean_ssn))),
  list(type = "boolean12", fun = list(naaccr_boolean12), fun_unknown = list(naaccr_boolean12)),
  list(type = "Date", fun = list(naaccr_date), fun_unknown = list(naaccr_date)),
  list(type = "datetime", fun = list(naaccr_datetime), fun_unknown = list(naaccr_datetime))
))


#' Define custom fields for NAACCR records
#'
#' Create a \code{record_format} object, which is used to read NAACCR records.
#'
#' To define registry-specific fields in addition to the standard fields, create
#' a \code{record_format} object for the registry-specific fields and combine it
#' with one of the formats provided with the package using \code{rbind}.
#'
#' @param name Item name appropriate for a \code{data.frame} column name.
#' @param item NAACCR item number.
#' @param start_col First column of the field in a fixed-width record.
#' @param end_col *Deprecated: Use the \code{width} parameter instead.*
#'   Last column of the field in a fixed-width record.
#' @param type Name of the column class.
#' @param alignment Alignment of the field in fixed-width files. Either
#'   \code{"left"} (default) or \code{"right"}.
#' @param padding Single-character strings to use for padding in fixed-width
#'   files.
#' @param parent Name of the parent node to include this field under when
#'   writing to an XML file.
#'   Values can be \code{"NaaccrData"}, \code{"Patient"}, \code{"Tumor"}, or
#'   \code{NA} (default).
#'   Fields with \code{NA} for parent won't be included in an XML file.
#' @param cleaner (Optional) List of functions to handle special cases of
#'   cleaning field data (e.g., convert all values to uppercase).
#'   Values of \code{NULL} (the default) mean the default cleaning function for
#'   the \code{type} is used.
#'   The value can also be the name of a function to retrieve with
#'   \code{\link[methods:methodUtilities]{getFunction}}.
#'   See Details.
#' @param unknown_finder (Optional) List of functions to detect when codes mean
#'   the actual values are unknown or not applicable.
#'   Values of \code{NULL} (the default) mean the default unknown finding
#'   function for the \code{type} is used.
#'   The value can also be the name of a function to retrieve with
#'   \code{\link[methods:methodUtilities]{getFunction}}.
#'   See Details.
#' @param name_literal (Optional) Item name in plain language.
#' @param width (Optional) Item width in characters.
#' @param x Object to be coerced to a \code{record_format}, usually a
#'   \code{data.frame} or \code{list}.
#' @param ... Other arguments passed to \code{record_format}.
#'
#' @return An object of class \code{"record_format"} which has the following
#'   columns:
#'   \describe{
#'     \item{\code{name}}{
#'       (\code{character}) XML field name.
#'     }
#'     \item{\code{item}}{
#'       (\code{integer}) Field item number.
#'     }
#'     \item{\code{start_col}}{
#'       (\code{integer}) First column of the field in a fixed-width text file.
#'       If \code{NA}, the field will not be read from or written to fixed-width
#'       files. They will included in XML files.
#'     }
#'     \item{\code{end_col}}{
#'       (\code{integer}) (*Deprecated: Use \code{width} instead.*)
#'       Last column of the field in a fixed-width text file.
#'       If \code{NA}, the field will not be read from or written to fixed-width
#'       files. This is the norm for fields only found in XML formats.
#'     }
#'     \item{\code{type}}{
#'       (\code{factor}) R class for the column vector.
#'     }
#'     \item{\code{alignment}}{
#'       (\code{factor}) Alignment of the field's values in a fixed-width
#'       text file.
#'     }
#'     \item{\code{padding}}{
#'       (\code{character}) String used for padding field values in a
#'       fixed-width text file.
#'     }
#'     \item{\code{parent}}{
#'       (\code{factor}) Parent XML node for the field. One of
#'       \code{"NaaccrData"}, \code{"Patient"}, or \code{"Tumor"}.
#'     }
#'     \item{\code{cleaner}}{
#'       (\code{list} of \code{function} objects) Function to prepare the
#'       field's values for analysis.
#'       Values of \code{NULL} will use the standard cleaner functions for the
#'       \code{type} (see below).
#'     }
#'     \item{\code{unknown_finder}}{
#'       (\code{list} of \code{function} objects) Function to detect codes
#'       meaning the actual values are missing or unknown for the field.
#'     }
#'     \item{\code{name_literal}}{
#'       (\code{character}) Field name in plain language.
#'     }
#'     \item{\code{width}}{
#'       (\code{integer}) Character width of the field values.
#'       Mostly meant for reading and writing flat files.
#'     }
#'   }
#'
#' @section Format Types:
#'
#'   The levels \code{type} can take, along with the functions used to process
#'   them when reading a file:
#'
#'   \describe{
#'     \item{\code{address}}{
#'       (\code{\link{clean_address_number_and_street}})
#'       Street number and street name parts of an address.
#'     }
#'     \item{\code{age}}{
#'       (\code{\link{clean_age}})
#'       Age in years.
#'     }
#'     \item{\code{boolean01}}{
#'       (\code{\link{naaccr_boolean}}, with \code{false_value = "0"})
#'       True/false, where \code{"0"} means false and \code{"1"} means true.
#'     }
#'     \item{\code{boolean12}}{
#'       (\code{\link{naaccr_boolean}}, with \code{false_value = "1"})
#'       True/false, where \code{"1"} means false and \code{"2"} means true.
#'     }
#'     \item{\code{census_block}}{
#'       (\code{\link{clean_census_block}})
#'       Census Block ID number.
#'     }
#'     \item{\code{census_tract}}{
#'       (\code{\link{clean_census_tract}})
#'       Census Tract ID number.
#'     }
#'     \item{\code{character}}{
#'       (\code{\link{clean_text}})
#'       Miscellaneous text.
#'     }
#'     \item{\code{city}}{
#'       (\code{\link{clean_address_city}})
#'       City name.
#'     }
#'     \item{\code{count}}{
#'       (\code{\link{clean_count}})
#'       Integer count.
#'     }
#'     \item{\code{county}}{
#'       (\code{\link{clean_county_fips}})
#'       County FIPS code.
#'     }
#'     \item{\code{Date}}{
#'       (\code{\link{as.Date}}, with \code{format = "\%Y\%m\%d"})
#'       NAACCR-formatted date (YYYYMMDD).
#'     }
#'     \item{\code{datetime}}{
#'       (\code{\link{as.POSIXct}}, with \code{format = "\%Y\%m\%d\%H\%M\%S"})
#'       NAACCR-formatted datetime (YYYYMMDDHHMMSS)
#'     }
#'     \item{\code{facility}}{
#'       (\code{\link{clean_facility_id}})
#'       Facility ID number.
#'     }
#'     \item{\code{icd_9}}{
#'       (\code{\link{clean_icd_9_cm}})
#'       ICD-9-CM code.
#'     }
#'     \item{\code{icd_code}}{
#'       (\code{\link{clean_icd_code}})
#'       ICD-9 or ICD-10 code.
#'     }
#'     \item{\code{integer}}{
#'       (\code{\link{as.integer}})
#'       Miscellaneous whole number.
#'     }
#'     \item{\code{numeric}}{
#'       (\code{\link{as.numeric}})
#'       Miscellaneous decimal number.
#'     }
#'     \item{\code{override}}{
#'       (\code{\link{naaccr_override}})
#'       Field describing why another field's value was over-ridden.
#'     }
#'     \item{\code{physician}}{
#'       (\code{\link{clean_physician_id}})
#'       Physician ID number.
#'     }
#'     \item{\code{postal}}{
#'       (\code{\link{clean_postal}})
#'       Postal code for an address (a.k.a. ZIP code in the United States).
#'     }
#'     \item{\code{ssn}}{
#'       (\code{\link{clean_ssn}})
#'       Social Security Number.
#'     }
#'     \item{\code{telephone}}{
#'       (\code{\link{clean_telephone}})
#'       10-digit telephone number.
#'     }
#'   }
#'
#' @examples
#'   my_fields <- record_format(
#'     name      = c("foo", "bar", "baz"),
#'     item      = c(2163, 1180, 1181),
#'     start_col = c(975, 1381, NA),
#'     width     = c(1, 55, 4),
#'     type      = c("numeric", "facility", "character"),
#'     parent    = c("Patient", "Tumor", "Tumor"),
#'     cleaner   = list(NULL, NULL, trimws)
#'   )
#'   my_format <- rbind(naaccr_format_16, my_fields)
#' @import data.table
#' @export
record_format <- function(name,
                          item,
                          start_col = NA_integer_,
                          end_col = NA_integer_,
                          type = "character",
                          alignment = "left",
                          padding = " ",
                          parent = "Tumor",
                          cleaner = list(NULL),
                          unknown_finder = list(NULL),
                          name_literal = NA_character_,
                          width = NA_integer_) {
  if (!identical(end_col, NA_integer_) && identical(width, NA_integer_)) {
    warning("end_col is deprecated. Use the width paramter instead.")
  }
  if (is.function(cleaner)) {
    cleaner <- list(cleaner)
  } else if (is.atomic(cleaner)) {
    cleaner <- as.list(as.character(cleaner))
  }
  if (is.function(unknown_finder)) {
    unknown_finder <- list(unknown_finder)
  } else if (is.atomic(unknown_finder)) {
    unknown_finder <- as.list(as.character(unknown_finder))
  }
  # Allow 0-row formats, because why not?
  if (length(name) == 0L && length(item) == 0L) {
    start_col <- start_col[0L]
    end_col <- end_col[0L]
    type <- type[0L]
    alignment <- alignment[0L]
    padding <- padding[0L]
    parent <- parent[0L]
    cleaner <- cleaner[0L]
    unknown_finder <- unknown_finder[0L]
    name_literal <- name_literal[0L]
    width <- width[0L]
  }
  padding   <- as.character(padding)
  padding_width <- nchar(padding)
  if (any(padding_width > 1L, na.rm = TRUE)) {
    stop("'padding' must only contain single-character values")
  }
  parent_nodes <- c("NaaccrData", "Patient", "Tumor")
  if (!all(parent %in% c(NA, parent_nodes))) {
    parent_list <- paste0("'", parent_nodes, "'", collapse = ", ")
    warning("Replacing values of 'parent' other than (", parent_list, ") with NA")
  }
  # Create the format
  fmt <- data.table(
    name = as.character(name),
    item = as.integer(item),
    start_col = as.integer(start_col),
    end_col = as.integer(end_col),
    type = factor(as.character(type), sort(type_converters[["type"]])),
    alignment = factor(as.character(alignment), c("left", "right")),
    padding = as.character(padding),
    parent = factor(parent, parent_nodes),
    cleaner = as.list(cleaner),
    unknown_finder = as.list(unknown_finder),
    name_literal = as.character(name_literal),
    width = as.integer(width)
  )
  inferred_ends <- fmt[["start_col"]] + fmt[["width"]] - 1L
  if (any(fmt[["end_col"]] != inferred_ends, na.rm = TRUE)) {
    warning("Some end_col values have been replaced using start_col and width.")
  }
  # Replace mismatched values and NA
  fmt[["end_col"]] <- inferred_ends
  if (anyNA(fmt[["alignment"]])) {
    stop("'alignment' must only contain values of \"left\" or \"right\"")
  }
  if (anyNA(fmt[["type"]])) {
    stop(
      "'type' must be one of ",
      paste0("'", levels(fmt[["type"]]), "'", collapse = ", ")
    )
  }
  setattr(fmt, "class", c("record_format", class(fmt)))
  fmt
}


#' @rdname record_format
#' @importFrom utils modifyList
#' @export
as.record_format <- function(x, ...) {
  if (inherits(x, "record_format")) {
    return(x)
  }
  xlist <- as.list(x)
  xlist <- utils::modifyList(xlist, list(...), keep.null = TRUE)
  if (all(c("width", "end_col") %in% names(xlist))) {
    xlist[["end_col"]] <- NULL
  }
  call_args <- args(record_format)
  arg_names <- intersect(names(xlist), names(as.list(call_args)))
  arg_names <- arg_names[nzchar(arg_names)]
  do.call(record_format, xlist[arg_names])
}


#' @noRd
rbind.record_format <- function(..., stringsAsFactors = FALSE) {
  formats <- lapply(list(...), as.record_format)
  combined <- rbindlist(formats)
  as.record_format(combined)
}


#' Field definitions from all NAACCR format versions
#'
#' Each \code{naaccr_format_XX} object is a \code{data.table} defining the
#' fields for each version of NAACCR's record file format.
#' \code{naaccr_formats} is a list of these record formats, with each name
#' being the two- or three-digit code for the format.
#'
#' @description See \code{\link{record_format}}.
#'
#' @rdname naaccr_formats
#' @export
"naaccr_formats"

#' @rdname naaccr_formats
#' @export
"naaccr_format_12"

#' @rdname naaccr_formats
#' @export
"naaccr_format_13"

#' @rdname naaccr_formats
#' @export
"naaccr_format_14"

#' @rdname naaccr_formats
#' @export
"naaccr_format_15"

#' @rdname naaccr_formats
#' @export
"naaccr_format_16"

#' @rdname naaccr_formats
#' @export
"naaccr_format_18"

#' @rdname naaccr_formats
#' @export
"naaccr_format_21"

#' @rdname naaccr_formats
#' @export
"naaccr_format_22"

#' @rdname naaccr_formats
#' @export
"naaccr_format_23"


#' Internal function for other functions to resolve format
#' @noRd
choose_naaccr_format <- function(version = NULL, format = NULL, keep_fields = NULL) {
  if (is.null(version) && is.null(format)) {
    version <- max(names(naaccr_formats))
  } else if (!is.null(version) && !is.null(format)) {
    stop("Specify 'version' or 'format', not both")
  }
  if (!is.null(version)) {
    version <- formatC(as.integer(version), format = "d")
    fmt <- naaccr_formats[[version]]
    if (is.null(fmt)) {
      valid_versions <- names(naaccr_formats)
      stop(
        "version must be a valid NAACCR format version number from: ",
        paste0(valid_versions, collapse = ", ")
      )
    }
  } else {
    fmt <- as.record_format(format)
  }
  if (!is.null(keep_fields)) {
    fmt <- fmt[list(name = as.character(keep_fields)), on = "name"]
  }
  as.record_format(fmt)
}