---
title: "Package structure"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{structure}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#"
)
```

# Purpose

This vignette is aimed at developers who want to understand the package better and to make it easier for them to contribute.

# Overview

There are only two main user-facing functions in {a11ytables}: 

* `create_a11ytable()` to create a data.frame object (with an additional 'a11ytable' S3 class) filled with all the information needed to create a spreadsheet output, as well as check the validity of the structure and provide errors or warnings
* `generate_workbook()` to convert the output from `create_a11ytable()` to an [{openxlsx}](https://ycphs.github.io/openxlsx/) Workbook-class object, ready for writing to an xlsx file with `openxlsx::saveWorkbook()`

This simplicity is a feature, not a bug. It's designed to greatly simplify the process of creating compliant spreadsheets. The package does the hard work of making the outputs compliant so the user spends less time dealing with it.

This vignette provides a quick look at what's happening 'under the hood' in these functions.

Please add [an issue](https://github.com/co-analysis/a11ytables/issues) to the package's GitHub repository if you would like any of this explanation to be expanded, or provide a solution in [a pull request](https://github.com/co-analysis/a11ytables/pulls).

# Files

First it's worth explaining how the source files are laid out. There are four major groups of scripts in the `R/` directory of the package:

1. Code to make a11ytable-class objects: `a11ytable.R` and `utils-a11ytable.R` contain code for handling the a11ytable class, most importantly the `create_a11ytable()` function, but also coercion with `as_a11ytable()`, checking with `is_a11ytable()`, a `summary()` method and a `print()` method, which takes advantage of [the {pillar} package](https://github.com/r-lib/pillar) for prettier outputs.
2. Code to make Workbook-class objects: `workbook.R`, `utils-workbook.R` and `utils-workbook-style.R` contain the code for creating and styling a Workbook-class object with the `generate_workbook()` function.
3. Code to produce demo datasets: `data.R` contains the documentation for demo datasets, which are created in the `data-raw/` directory with the files stored in the `data/` directory.
4. Code that creates the RStudio Addin: `addin.R` and `utils-addin.R` contain code for the [RStudio Addin](https://rstudio.github.io/rstudioaddins/) (the .dcf file for which is in the `inst/rstudio/` directory).

You'll also find the `a11ytables-package.R` file in the `R/` directory, which provides a package-level help page derived from the DESCRIPTION file when `?a11ytables` is run by the user. It doesn't need to be edited.

# Code

This sections below focus on the `create_a11ytable()` and `generate_workbook()` functions, which are the primary and most complex functions in the package.

The code that underpins these functions is modularised to aid with bug-catching and testing, but also to make it easier for developers to understand how the code fits together. Internal sub-functions are consistently-named and begin with verbs, which should help you better understand their purpose.

Note that {a11ytables} uses a convention that internal functions (i.e. those not presented to the user, but accessed via the `:::` qualifier) are prefixed with a period (i.e. `.f()`) to make it clearer that they are internal to the package. The exported user-facing functions do not use a leading period.

## To create a11ytables

Actually, `create_a11ytable()` itself only does one thing: it takes user inputs from the arguments and combines them into a dataframe. It then passes this off to the most important function in the package, `as_a11ytable()`, which is responsible for coercing the dataframe to a11ytable class and performing checks on its content.

Basically, `as_a11ytable()` creates an S3-class object with classes 'data.frame' and 'tbl' (i.e. [tibble](https://tibble.tidyverse.org/)) and an additional 'a11ytable' class.

```{r class-demo}
library(a11ytables)
my_a11ytable <- as_a11ytable(demo_df)
class(my_a11ytable)
```

The object can be manipulated like a 'normal' dataframe and---thanks to the {pillar} package and the tbl class---it can be printed in compact form without the need for the whole of the {tibble} package to be imported.

```{r print-demo}
my_a11ytable
```

Compare this to its appearance as a regular data.frame, which is trickier to understand:

```{r df-demo-no-eval}
as.data.frame(my_a11ytable)
```

Within `as_a11ytable()` itself are two major functions that help ensure proper construction of an a11ytable object:

* `.validate_a11ytable()` will generate errors if basic structural expectations of an a11ytable aren't met (e.g. if 'cover', 'contents' or 'notes' have been provided more than once to the `sheet_type` argument)
* `.warn_a11ytable()` checks for things that the user may have forgotten and prints warnings about them (e.g. if 5 notes are declared in the notes sheet but there are fewer in the tables themselves)

Advanced users can create a correctly-formatted data.frame on the fly and convert it to an a11ytable with `as_a11ytable()` directly. The `as_a11ytable()` function mainly exists to make testing easier, i.e. you can pass to it the pre-prepared `demo_df` dataset.

### Methods

There's a few methods for a11ytables that are also found in `R/a11ytables.R`.

`is_a11ytable()` is a classic logical test that checks for the a11ytable class in the object provided to it.

```{r is-demo}
is_a11ytable(my_a11ytable)
```

The `summary()` method prints a very simple overview of a provided a11ytable.

```{r summary-demo}
summary(my_a11ytable)
```

The `tbl_sum()` method is provided via the {pillar} package, with the goal of providing a bespoke header to the printed a11ytable.

```{r tbl-sum-demo}
pillar::tbl_sum(my_a11ytable)
```

## To create workbooks

The `generate_workbook()` function sets up an [{openxlsx}](https://ycphs.github.io/openxlsx/) Workbook-class object and fills it by iterating over a user-supplied the a11ytable-class object.

```{r wb-class}
my_wb <- generate_workbook(my_a11ytable)
class(my_wb)
```

You can see how the Workbook-class object carries information that will determine the structure and style of the final spreadsheet output.

```{r print-wb}
my_wb
```

Several internal sub-functions within `generate_workbook()`---`.add_*()`, `.insert_*()` and `.style_*()`---are responsible for adding these sheets, inserting sheet elements and styling them, respectively.

### Add sheets

A Workbook-class object is first created with `openxlsx::createWorkbook()` and then sheets are added based on the contents of the user-supplied a11ytable. 

The following functions add sheets and sheet elements into the workbook:

* `.add_tabs()` adds the required number of tabs into the workbook with `openxlsx::addWorksheet()` (as per the `tab_title` column of the supplied a11ytable)
* `.add_cover()` and `.add_contents()` add the information needed for the cover and contents sheets (as per the required 'cover' and 'contents' supplied in the `sheet_type` column of an a11ytable)
* `.add_notes()` if a notes sheet exists (i.e. a row in the supplied a11ytable with a `sheet_type` of 'notes')
* `.add_table()` adds sheets for each statistical table (as per rows of supplied a11ytable with a `sheet_type` of 'table')

As sheets are added, content is inserted and styles are applied with the:

* `.insert_*()` functions, which insert sheet elements (title, source statement, table, etc) to each sheet
* `.style_*()` functions, which apply formatting to each sheet (e.g. bold sheet titles with larger font) and the workbook (e.g. Arial font)

#### Insert sheet elements

There are several `.insert_*()` functions that add information to each sheet depending on the `sheet_type` of the provided a11ytable, as well as the content, if any, of its `sheet_title`, `blank_cells`, `source` and `table` columns.

The following functions insert 'pre-table' elements in this order:

* `.insert_title()` to place the sheet title in cell A1
* `.insert_table_count()` to add a statement about the number of tables in the sheet
* `.insert_notes_statement()` if a `sheet_type` of 'notes' is provided in the user's a11ytable
* `.insert_blanks_message()` if content is provided in the `blanks_cells` column of the user's a11ytable
* `.insert_custom_rows()` if content is provided in the `custom_rows` column of the user's a11ytable
* `.insert_source()` if content is provided in the `source` column of the user's a11ytable
    
A table of data is added under the metadata with `.insert_table()`, which is provided in the `table` column of the user's a11ytable object.

The exact `.insert_*()` functions called depend on the `sheet_type` declared in the a11ytable:

* meta sheets (cover, contents and notes) need only `.insert_title()` and `.insert_table_count()`
* statistical tables may also require `.insert_blanks_message()`, `.insert_custom_rows()` and `.insert_source()` if the relevant content is provided by the user, as well as `.insert_notes_statement()` if there are notes

Simple logic is used to check for the presence of meta elements with the `.has_*()` functions, while the `get_start_row_*()` functions handle the cell to which each message should be inserted. 

For example, if all the elements are supplied, then the table would begin in row 6 (i.e. after the sheet title, table count, note presence, meaning of blank cells and source), but it's possible that the table would have to be inserted to row 3 if only the sheet title and statement are required. This avoids inaccessible blank rows and redundant statements like 'This table has no source statement'.

#### Apply styles

There are a few `.style_*()` functions that create styles and apply them on the basis of the `sheet_type` provided in the a11ytable.

* `.style_create()` creates an easily-referenced lookup of styles, which is created with `openxlsx::createStyle()`
* `.style_workbook()` applies defaults for _the whole workbook_ (i.e. to set the font style to Arial size 12)
* `.style_cover()`, `.style_contents()` and `.style_notes()` all apply styles to specific _sheets_
* `.style_sheet_title()` and `.style_table()` apply styles to _particular sheet elements_ (e.g. the title is larger and bolder than the default font)

# Contribute

To contribute, please add [an issue](https://github.com/co-analysis/a11ytables/issues) or [a pull request](https://github.com/co-analysis/a11ytables/pulls) after reading [the code of conduct](https://github.com/co-analysis/a11ytables/blob/main/CODE_OF_CONDUCT.md) and [contributing](https://github.com/co-analysis/a11ytables/blob/main/.github/CONTRIBUTING.md) guidance.