---
title: "Get started with tabular"
vignette: >
  %\VignetteIndexEntry{Get started with tabular}
  %\VignetteEngine{quarto::html}
  %\VignetteEncoding{UTF-8}
---

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

`tabular` turns one **pre-summarised, wide data frame** into a publication-ready
table and renders it natively to **RTF, HTML, DOCX, PDF/LaTeX and Markdown** —
from a single spec, with no Java or Office dependency.

Two things to internalise up front:

1. **tabular is display-only.** It never aggregates, filters, or computes
   statistics. You bring a summarised data frame (one input row = one display
   row); tabular lays it out and renders it. (Producing that frame from a cards
   ARD is the [*Data in*](https://vthanik.github.io/tabular/articles/data-in.html) article.)
2. **One immutable spec, built with verbs.** You pipe a `tabular()` object
   through verbs (`cols()`, `headers()`, `paginate()`, `preset()`, …) and finish
   with `emit()`. Each verb returns a new spec; nothing renders until `emit()`.

## Your first table

Start from a wide frame — here the bundled demographics summary, one row per
statistic, one column per treatment arm:

```{r}
data(cdisc_saf_demo, package = "tabular")
head(cdisc_saf_demo)
```

Describe the columns. The spec prints as a live HTML table — this is the same
render `emit()` produces, shown inline:

```{r}
spec <- tabular(
  cdisc_saf_demo,
  titles = c(
    "Table 14-2.01",
    "Demographic and Baseline Characteristics",
    "ITT Population"
  )
) |>
  cols(
    variable = col_spec(
      usage = "group",
      group_display = "header_row",
      label = ""
    ),
    stat_label = col_spec(label = "")
  )

spec
```

To write a file, hand the spec to `emit()`; the backend is chosen by the file
extension (or an explicit `format =`):

```{r}
out <- tempfile(fileext = ".rtf")
emit(spec, out) # RTF here; swap to .docx / .pdf / .html / .md
file.exists(out)
```

That is the whole loop: **wide frame → `tabular()` → `cols()` → `emit()`** — one
spec, any backend.

## The pipeline at a glance

Read it left to right. You **summarise upstream** — with cards/cardx, dplyr, or
SAS — into a long ARD, widen that to a display-ready frame with `pivot_across()`,
then hand it to tabular. Inside the package the work happens in three phases
(**Build → Resolve → Emit**), and the *same* resolved spec emits to every
backend, so the HTML you preview and the RTF you ship can never disagree.

![tabular's pipeline: summarise upstream into a long ARD, widen it with `pivot_across()`, then build, resolve, and emit one immutable spec to every backend.](../man/figures/workflow.svg){fig-alt="An ADaM dataset is aggregated into a long ARD, widened by pivot_across into a wide data frame, then Build, Resolve, and Emit render that one spec to RTF, PDF, HTML, LaTeX, and DOCX." width="100%"}

## Anatomy of a clinical table page

A submission table is not just a grid of numbers — it is a page with four stacked
sections, and a reviewer expects each one in its place. Every tabular verb maps
onto a piece of this picture:

![The four-section clinical page and the verb that fills each section.](../man/figures/anatomy.svg){fig-alt="A clinical table page split into header section, title lines, data section, and footnote lines, each annotated with the tabular verb that produces it."}

- **Header section** — the running protocol, optional status, and *page x of y*,
  set as page chrome with `preset(pagehead =, pagefoot =)`.
- **Title lines** — the table number and up to four centred titles, passed to
  `tabular(titles =)`.
- **Data section** — an optional `subgroup()` banner, the column-header band built
  by `cols()` and `headers()`, then the decimal-aligned data.
- **Footnote lines** — your static `footnotes =` plus any auto-numbered
  `footnote()` markers, then the program path, name, and timestamp.

`preset()` frames all four by controlling the page geometry (paper, orientation,
margins, fonts).

## Where to next

The rest of the docs are task-oriented — read the one that matches what you are
doing:

- **[Data in](https://vthanik.github.io/tabular/articles/data-in.html)** — turn a cards/cardx ARD (or any long ARD) into
  the wide frame, with `pivot_across()`.
- **[Structure](https://vthanik.github.io/tabular/articles/structure.html)** — columns, headers, BigN, and splitting wide
  or long tables across pages (`cols()`, `headers()`, `paginate()`,
  `subgroup()`).
- **[Presentation](https://vthanik.github.io/tabular/articles/presentation.html)** — titles, footnotes, running headers,
  and cell styling (`footnote()`, `preset()`, `style()`).
- **[Recipes](https://vthanik.github.io/tabular/articles/recipes.html)** — the canonical CDISC-pilot safety and efficacy
  tables built end to end, each rendered live.
- **[Output & qualification](https://vthanik.github.io/tabular/articles/output.html)** — the backends, their system
  requirements, and the CDISC-pilot cross-backend validation.