Maintenance guidance

2021-03-14

Intoduction

DescrTab2 is a powerful package with vast customization options. With this, unfortunately, comes code that has to deal with quite a bit of special cases an exceptions. This document aims to describe the flow of control of the DescrTab2 package, so that future generations may continue development and successfully fix potential bugs.

Flow of control

descr

The user interfaces mostly with the descr function. descr does all the calculations, i.e. the evaluation of the summary statistics on the data (mean, sd, median, etc. for continuous variables and counts for categorical variables) and the calculation of statistical tests. For this descr calls the descr_cat and descr_cont functions, which evaluate the list of summary statistics on the data. descr_cat then calls test_cat and descr_cont calls test_cont, which calculate appropriate statistical tests. A detailed description for the choice of test can be read in the “Test choice” vignette.

descr returns a DescrList object, which is basically a named list containing all calculation results and the formatting options.

print

To turn a DescrList object into pretty output, the object has to be passed to the print function. print is a generic function. This means that if a DescrList object is passed to print, the specialized print.DescrList function will be invoked automatically.

Preprocessing

Since the proper output format is highly document type dependent, print.DescrList creates output in two steps. The first step is independent of the output format: The creation of a DescrPrintObj by calling the create_printObj function.

In this function, proper formatting is applied to the results in the DescrList and the formatted values are saved inside a tibble. Formatting in this case means converting numbers to characters, reducing the number of decimal digits, combining variables like “Q1” and “Q3” into “Q1 - Q3”, formatting small p values to display as “<0.001” and adding “%” values to categorical variables.

Somewhat of an exception is the case printFormat="numeric". Here, numbers are not converted characters and consequently very little formatting can be applied.

The formatting in create_printObj is done by iterating over all variables in the DescrList object and creating an appropriate sub-table by calling one of create_numeric_subtable.cat_summary, create_numeric_subtable.cont_summary, create_character_subtable.cat_summary or create_character_subtable.cont_summary. Whether create_numeric_subtable or create_character_subtable is called is determined by the printFormat option (all options lead to create_character_subtable except printFormat="numeric). Whether .cat_summary or .cont_summary is called depends on the type of variable. The sub-tables are then concatenated to a master table.

Postprocessing

The DescrPintObj is the transformed into appropriate output format by calling one of print_tex, print_html, print_word, print_console or print_numeric.

print_console basically prints the tibble that is produces by create_printObj using a slightly modified version of the default method for printing tibbles.

print_numeric basically prints the tibble produces by create_printObj if printFormat="numeric" was specified.

print_tex and print_html use kableExtra to convert the tibble from create_printObj into raw tex or html output. Some special formatting has to be applied to these outputs to accomodate for superscripts and to escape special LaTeX characters.

print_word produces a flextable object from the tibble returned by create_printObj. flextables play relatively nicely with word.