---
title: "Interoperability and export"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Interoperability and export}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(collapse = TRUE, comment = "#>")
has_sfn <- requireNamespace("sfnetworks", quietly = TRUE)
has_tg  <- requireNamespace("tidygraph", quietly = TRUE)
```

```{r setup}
library(osmnxr)
g <- ox_example("olinda") # a small real network bundled with the package
```

An `osm_graph` is deliberately thin: tidy `sf` nodes and edges. That makes it
easy to hand off to the rest of the R spatial-network ecosystem, or to export
for other tools.

## sf

The nodes and edges are always available as `sf`:

```{r}
parts <- ox_as_sf(g)
parts$edges
```

## sfnetworks and tidygraph

Convert to an `sfnetwork` for the `tidygraph` verb workflow, or to a bare
`tbl_graph`:

```{r, eval = has_sfn}
net <- ox_as_sfnetwork(g)
net
```

```{r, eval = has_tg}
tg <- ox_as_tidygraph(g)
tg
```

## dodgr

`ox_as_dodgr()` returns the column layout `dodgr` expects, so you can use its
highly optimised routing directly:

```{r, eval = FALSE}
library(dodgr)
graph <- ox_as_dodgr(g)
dodgr_dists(graph, from = graph$from_id[1], to = graph$to_id[10])
```

## GeoJSON and MapLibre

Export edges to GeoJSON, or build a MapLibre GL style fragment that points at
the written data:

```{r}
gj <- tempfile(fileext = ".geojson")
ox_to_geojson(g, gj)

style <- ox_to_maplibre(g, tempfile(fileext = ".geojson"))
str(style, max.level = 2)
```

## GraphML round-trip

`ox_save_graphml()` / `ox_load_graphml()` persist the graph in a format
compatible with OSMnx, NetworkX and Gephi. Edge geometry is stored as WKT, so
the round-trip is lossless:

```{r}
f <- tempfile(fileext = ".graphml")
ox_save_graphml(g, f)
g2 <- ox_load_graphml(f)
ox_basic_stats(g2)
```