Skip to content

Latest commit

 

History

History
82 lines (65 loc) · 3.84 KB

new_fn_notes.md

File metadata and controls

82 lines (65 loc) · 3.84 KB
Raw

New function notes for jamba

  • grepls1() - easy extension of grepls(..., 1) which searches only the user environment.

Text-centric, or console-centric

  • plotSmoothScatterText() - text version of plotSmoothScatter() that uses package "nostalgiR" enhanced with density per point. Optional ability to tile multiple plot panels, stopping short of replicating ggplot2.
  • imageByColorsText() - text version of imageByColors() which takes a matrix of colors, a matrix of labels, and displays it as an image. Bonus points for centering text labels across multiple neighboring cells with same color and label.

colortable()

A long-awaited port of a function I previously used all the time but have not properly ported into an R package, and so I miss using it. sad face

  • It trims displayed column width so all columns fit the console width

  • Columns are numbered, rows are numbered.

  • Highlighted colnames and rownames

  • Row names are highlighted, encouraged, and warmly embraced. (Sorry tibble!)

  • Numeric values are formatted for friendly output:

    • values above 1000 use big.mark="," to appear like this: 1,000
    • decimal values use reasonable significant digits, so instead of displaying 3.141592653589793116 it will display 3.14.
    • however columns with values c(1000, 0.001) will be displayed c("1,000", "0.001") - and not the default which is to apply the same level of detail to every value, for example not this: c("1e+03", "1e-03"), and not this: c("1,000.000", "0.001")
    • P-values may have specific formatting rules, I forget the details rn.

  • If not all columns will fit the console width when shrunk down to some a minimum width, it splits output across multiple lines, and includes row name and row number beside each split.

    • For me, it is not helpful to print only the first n columns, the main purpose is to see every column.
    • For me, it is not helpful to split super-wide columns across multiple lines, sometimes this process splits into one chunk per column. For me it is much more helpful to shorten the displayed value in each cell.
    • There is some logic to determine a reasonable size to shrink each column, for example nchar() upper quartile.

  • Possibly the most useful feature is that values in each column are colorized. This step helps recognize outliers, typos, inconsistencies, etc.

  • Each column is colorized based upon its data type:

    • character columns are colorized using supplied color_sub for name:color substitution, otherwise it calls colorjam::rainbowJam().

    • numeric columns are colorized one of a few ways:

      • Default: divergent blue-white-red colors using colorjam::col_div_xf() extended to the maximum absolute value, linear scale. Option to apply quantile or percent-max logic to redefine the maximum value. Option to apply in log scale using jamba::log2signed().
      • P-values: colorized using -log10(x) which emphasizes P-values near zero, and therefore values near 1 are not colorized. Ideally, P-value columns are recognized by colname and/or by the numeric values.
      • Correlation: Same as Default but uses a fixed maximum=1 for consistency.
      • Global and per-column max values can be defined, to help apply a useful color gradient.

  • Columns can be sorted, by proxy of calling jamba::mixedSortDF() to enable mixed alphanumeric sorting.

  • Features intended but not enabled (yet):

    • optional word-wrapping column headers
    • optional word-wrapping character values, for up to n lines, expanding the number of lines used for affected rows.
    • Ability to view proper output from table(), and not the mangled tall-skinny format from as.data.frame(table(x))
    • Ability to view output from table() with more than two variables.
    • Ability to view output from n-dimensional arrays.