Capture melt"
In nc: Named Capture to Data Tables

Capture melt

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

This vignette explains how to use functions for "melting" wide data tables, i.e. converting to tall/long data tables. To clarify the discussion we introduce the following three terms:

Reshape columns contain the data which are present in equal amounts in the input and output, but in fewer columns in the output. These are measure.vars in reshape2/data.table.
Copy columns contain data which are copied from a single row of the input into several rows of the output. These are id.vars in reshape2/data.table.
Capture columns are only present in the output, and contain data extracted from the names of the input reshape columns, using a capturing regular expression.

The nc functions use data.table::melt internally:

nc::capture_melt_single is useful when you have a wide data table with lots of columns, all of the same type, that you would like to melt/reshape into a single result/output column. This uses melt with a character vector for measure.vars.
nc::capture_melt_multiple is useful when you have a wide data table with columns of possibly different types that you would like to melt/reshape into multiple result/output columns. This uses melt with a list for measure.vars.

Both are useful when you want to use a regular expression to specify both (1) the set of input columns to reshape and (2) some information to extract from those column names.

Reshaping several input columns into a single output column

Sometimes you want to melt a "wide" data table which has several distinct pieces of information encoded in each column name. One example is the familiar iris data, which have flower part and measurement dimension encoded in each of four column names:

head(iris)

Those four reshape column names can be specified via a regex in nc::capture_melt_single. The first argument is the input data table to reshape, and the subsequent arguments are interpreted as a pattern which is passed to nc::capture_first_vec. Any input column names which match the specified regex will be passed as measure.vars to melt:

(iris.tall <- nc::capture_melt_single(
  iris,
  part=".*",
  "[.]",
  dim=".*",
  value.name="cm"))

Note the output above has one copy column (Species), two capture columns (part, dim), and one reshape column (cm). Internally the function joins the result of nc::capture_first_vec (on column names) to the result of melt (on the data).

The reshaped data can be plotted with different parts on rows and different dimensions on columns:

if(require(ggplot2)){
  ggplot()+
    theme_bw()+
    theme(panel.spacing=grid::unit(0, "lines"))+
    facet_grid(part ~ dim)+
    geom_bar(aes(cm, fill=Species), data=iris.tall)
}

Reshaping several input columns into multiple output columns

We could instead use capture_melt_multiple to get multiple output columns. Like capture_melt_single, the first argument of capture_melt_multiple is the subject data table and the following arguments form a pattern which is matched to the input data column names. However the pattern must have at least two groups:

One group must be named column. This group is used to capture the part of the input column names which will be used for the output column names. The output will contain one column for each unique value captured in this group.
One or more other groups (with any names) specify an identifier for each observation.

(iris.part.cols <- nc::capture_melt_multiple(
  iris,
  column=".*",
  "[.]",
  dim=".*"))

Note that the reshaped table above contains one copy column (Species), one capture column (dim), and two reshape columns (Petal, Sepal). We can plot these data to see whether or not sepals are bigger than petals:

if(require(ggplot2)){
  ggplot()+
    theme_bw()+
    theme(panel.spacing=grid::unit(0, "lines"))+
    facet_grid(dim ~ Species)+
    coord_equal()+
    geom_abline(slope=1, intercept=0, color="grey")+
    geom_point(aes(
      Petal, Sepal),
      data=iris.part.cols)
}

It is clear from the plot above that sepals are indeed both longer and wider than petals, on each measured plant.

Melting WHO data with a more complex pattern

Another data set where it is useful to do column name pattern matching followed by melting is the World Health Organization data:

if(requireNamespace("tidyr")){
  data(who, package="tidyr")
}else{
  who <- data.frame(id=1, new_sp_m5564=2, newrel_f65=3)
}
names(who)

Each column which starts with new has three distinct pieces of information encoded in its name: diagnosis type (e.g. sp or rel), gender (m or f), and age range (e.g. 5564 or 1524). We would like to use a regex to match these column names, then using the matching columns as measure.vars in a melt, then join the two results.

new.diag.gender <- list(
  "new_?",
  diagnosis=".*",
  "_",
  gender=".")
nc::capture_melt_single(who, new.diag.gender, ages=".*")

Note the output includes the new reshape column called value by default, as in melt. The input reshape column names which matched the specified pattern, and there is a new column for each group in that pattern. The following example shows how to rename the value column and use numeric type conversion functions:

years.pattern <- list(new.diag.gender, ages=list(
  min.years="0|[0-9]{2}", as.numeric,
  max.years="[0-9]{0,2}", function(x)ifelse(x=="", Inf, as.numeric(x))))
(who.typed <- nc::capture_melt_single(
  who, years.pattern,
  value.name="count"))
str(who.typed)

Note in the code/result above that non-character captured output columns can be obtained by specifying type conversion functions in the pattern.

Melting monthly Eurostat data

Another example is exchange rate data from Eurostat. We first use read the entire data set into R:

ert.gz <- system.file(
  "extdata", "ert_eff_ic_m.tsv.gz", package="nc", mustWork=TRUE)
ert.all <- data.table::fread(ert.gz, na.strings=":")
ert.all[1:5, 1:5]

We see that the first column has some CSV data which we can parse via:

ert.first <- ert.all[, 1]
csv.lines <- c(sub("\\\\.*", "", names(ert.first)), ert.first[[1]])
ert.first.dt <- data.table::fread(text=paste(csv.lines, collapse="\n"))
ert.wide <- data.table::data.table(ert.first.dt, ert.all[,-1])
ert.wide[1:5, 1:5]

The wide data table can then be melted:

(ert.tall <- nc::capture_melt_single(
  ert.wide,
  year="[0-9]{4}", as.integer,
  "M",
  month="[0-9]{2}", as.integer))

After that we can create a time variable and plot via

ert.tall[, month.IDate := data.table::as.IDate(
  sprintf("%d-%d-15", year, month))]
if(require("ggplot2")){
  ggplot()+
    geom_hline(aes(
      yintercept=value),
      color="grey",
      data=data.frame(value=100))+
    geom_line(aes(
      month.IDate, value, color=geo),
      data=ert.tall[geo %in% c("CA", "US", "JP", "FR")])+
    facet_grid(exch_rt ~ .)+
    theme_bw()+
    theme(panel.spacing=grid::unit(0, "lines"))
}

Another way to do it would be via

nc::capture_melt_single(ert.wide, month.POSIXct="[0-9].*", function(x){
  as.POSIXct(strptime(paste0(x,"15"), "%YM%m%d"))
})

Melting into multiple output columns with missing input columns

What if the input data set has "missing" input columns?

iris.missing <- iris[, names(iris) != "Sepal.Length"]
head(iris.missing)

In that case melting into multiple columns is an error by default:

iris.pattern <- list(column=".*", "[.]", dim=".*")
nc::capture_melt_multiple(iris.missing, iris.pattern)

The error message explains that the number of input columns for each value of dim must be the same, but there is one for Length and two for Width. To ignore the error and fill the output with missing values,

nc::capture_melt_multiple(iris.missing, iris.pattern, fill=TRUE)

Note the missing values in the table above, which correspond to the missing input column in the original/wide data set.

Multiple output columns of different types

Some real-world data sets can be reshaped into output columns with different types. An example data set from the PROVEDIt benchmark in criminology:

peaks.csv <- system.file(
  "extdata", "RD12-0002_PP16HS_5sec_GM_F_1P.csv",
  package="nc", mustWork=TRUE)
peaks.wide <- data.table::fread(peaks.csv)
print(data.table::data.table(
  names=names(peaks.wide),
  class=sapply(peaks.wide, class)),
  topn=10)

There are 303 columns, with info for 100 peaks. Each peak has three features: Allele=character, Size=numeric, and Height=integer. The ending peaks are class logical because they are all missing. These data can be reshaped via

peaks.tall <- nc::capture_melt_multiple(
  peaks.wide,
  column=".*",
  " ",
  peak="[0-9]+", as.integer,
  na.rm=TRUE)
options(width=90)
print(peaks.tall)
str(peaks.tall)