Playing with webr – r.iresmi.net

A photo of an Impala browsing a tree — Impala (*Aepyceros melampus*) female browsing Weeping Boer-Bean (*Schotia brachypetala*) leaves – CC-BY-SA by Bernard DUPONT

webr allows R to run in the browser from a quarto HTML document. Let’s play!

Add webr to your quarto environment, in the terminal:

quarto add coatless/quarto-webr

Then, in the YAML header, add:

engine: knitr
filters: 
  - "webr"
webr:
  packages:
    - "dplyr"
    - "ggplot2"
    - "duckdb"
    - "glue"
    - "ggokabeito"

Now enclose your chunks with {webr-r} instead of {r}.

{dplyr} is a large package, so installing/loading it can take a little time…

… should work. You can edit and re-run the code in all chunks!

We will explore the popular french baby names. It is in parquet format; we can use {arrow} or {duckdb} to open it. Here we try with {duckdb}…

However we can’t INSTALL (yet) a DuckDB extension that would have allowed us to directly query over HTTP without downloading the whole file:

So we download and use a local copy of the dataset:

How does it look like?

Get the 2022 most popular names and check their past evolution:

Some were already very popular a long time ago. Half are recent. Names are now much more diluted.

--- title: "Playing with webr" description: "R in your browser" author: "Michaël" date: "2024-01-29T23:00:00" date-modified: last-modified categories: - R - webr - datavisualization draft: false freeze: true editor_options: chunk_output_type: console chunk_output_type: console execute: eval: true engine: knitr filters: ["webr"] webr: packages: - "ggplot2" - "dplyr" - "duckdb" - "glue" - "ggokabeito" --- ```{r} #| echo: false #| fig-cap: "Impala (*Aepyceros melampus*) female browsing Weeping Boer-Bean (*Schotia brachypetala*) leaves -- CC-BY-SA by Bernard DUPONT" #| fig-alt: "A photo of an Impala browsing a tree" #| fig-link: https://www.flickr.com/photos/berniedup/32415534737/ #| class-output: "preview-image" #| out-width: "100%" knitr::include_graphics("images/32415534737_1ebe274946_c.jpg") ``` [webr](https://github.com/coatless/quarto-webr) allows R to run in the browser from a [quarto](https://quarto.org/) HTML document. Let's play! Add webr to your quarto environment, in the terminal: ```{bash} #| eval: false quarto add coatless/quarto-webr ``` Then, in the YAML header, add: ```{yaml} #| eval: false engine: knitr filters: - "webr" webr: packages: - "dplyr" - "ggplot2" - "duckdb" - "glue" - "ggokabeito" ``` Now enclose your chunks with `{webr-r}` instead of `{r}`. {dplyr} is a large package, so installing/loading it can take a little time... ```{webr-r} mtcars |> filter(mpg > 30) ``` ... should work. You can edit and re-run the code in all chunks! We will explore the popular french baby names. It is in parquet format; we can use {arrow} or {duckdb} to open it. Here we try with {duckdb}... However we can't INSTALL (yet) a DuckDB extension that would have allowed us to directly query over HTTP without downloading the whole file: ```{webr-r} dataset <- "https://static.data.gouv.fr/resources/base-prenoms-insee-format-parquet/20231121-161435/prenoms-nat2022.parquet" cnx <- dbConnect(duckdb()) # To do once: dbExecute(cnx, "INSTALL httpfs") dbExecute(cnx, "LOAD httpfs") dbSendQuery(cnx, glue(" CREATE VIEW prenoms AS SELECT * FROM '{dataset}' WHERE annais <> 'XXXX'")) ``` So we download and use a local copy of the dataset: ```{webr-r} local_dataset <- "prenoms-nat2022.parquet" download.file(dataset, local_dataset) # create a duckdb view (dropping unknown years) dbSendQuery(cnx, glue(" CREATE VIEW prenoms AS SELECT * FROM '{local_dataset}' WHERE annais <> 'XXXX'")) ``` How does it look like? ```{webr-r} dbGetQuery(cnx, " DESCRIBE prenoms") dbGetQuery(cnx," SELECT COUNT(*) FROM prenoms") dbGetQuery(cnx," SELECT MIN(annais), MAX(annais) FROM prenoms") ``` Get the 2022 most popular names and check their past evolution: ```{webr-r} # how many names to select? n <- 9 top_2022 <- dbGetQuery(cnx, glue(" WITH top AS ( SELECT preusuel FROM prenoms WHERE annais = '2022' AND preusuel <> '_PRENOMS_RARES' ORDER BY nombre DESC LIMIT {n} ) SELECT p.preusuel, p.annais, p.nombre FROM prenoms p INNER JOIN top ON p.preusuel = top.preusuel")) head(top_2022) ``` ```{webr-r} #| label: fig-top-names #| fig-cap: "2022 popular names" #| fig-alt: !expr 'glue("Line plot of the top {n} 2022 popular names from 1900 to 2022")' #| out-width: "100%" #| fig-width: 7 top_2022 |> mutate(annais = as.integer(annais)) |> filter(nombre > 100) |> ggplot(aes(annais, nombre, color = preusuel)) + geom_line(lwd = .8) + scale_color_okabe_ito() + labs(title = glue("Top {n} 2022 names"), subtitle = "Most popular in France", x = "year", y = bquote(N[birth]), color = "names", caption = glue("données INSEE (icem7) https://www.data.gouv.fr/fr/datasets/base-prenoms-insee-format-parquet/ https://r.iresmi.net/ {Sys.Date()}")) + theme_minimal() + theme(plot.caption = element_text(size = 6), plot.background = element_rect(fill = "white")) ``` Some were already very popular a long time ago. Half are recent. Names are now much more diluted.