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

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

Discovery returns explicit identifiers. Pass that table to metadata or metrics;
the package handles batching and preserves your columns.

```{r, eval=FALSE}
st_publishers("Supercell") |>
  st_publisher_apps() |>
  mutate(cohort = "Supercell portfolio") |>
  st_metrics(date_from = "2026-01-01", date_to = "2026-03-31",
             countries = "US", granularity = "monthly")
```

Character IDs require `os`. Tables require `app_id` and `os`. Both native pipes
and magrittr pipes work. To change ID namespaces, use explicit provider mappings:

```{r, eval=FALSE}
st_apps("Clash of Clans") |>
  st_app(target_os = "ios") |>
  st_demographics(date_from = "2026-04-01", date_to = "2026-06-30")
```

There is no name-based merge. All mapped regional store IDs remain visible.
The `input_app_id` and `input_os` columns trace each mapping to its input.

Empty data stays typed and needs no token:

```{r}
tibble(app_id = character(), os = character(), cohort = character()) |>
  st_metrics(date_from = "2026-01-01", date_to = "2026-01-31")
```

Long metrics have explicit units and periods. DAU, WAU and MAU use native windows;
`granularity` changes sales aggregation only. Missing values are not zero. A
failed endpoint aborts unless you explicitly request `errors = "partial"`.
In partial mode, inspect `status`, `error` and `endpoint` before analysis.