Getting started with tidyprf • tidyprf

The Brazilian Federal Highway Police (Polícia Rodoviária Federal, PRF) publishes open data on traffic accidents and traffic violations on federal highways. tidyprf gives you these datasets directly in R, already cleaned and consolidated, with a consistent schema across all years.

library(tidyprf)

The three datasets

Dataset	Function	Unit	Available years
Accidents by person	`get_accidents()`	1 row per person involved	2007–2026
Accidents by occurrence	`get_crashes()`	1 row per accident	2007–2026
Traffic violations	`get_violations()`	1 row per violation	2019–2020, 2022–2026

Data are stored as one Parquet file per dataset per year, hosted on GitHub Releases. The first time you request a year, the file is downloaded and cached locally (in tools::R_user_dir("tidyprf", "cache")); subsequent calls read from the cache and require no internet connection.

Downloading data

All chunks in this section require an internet connection, so they are not evaluated when the vignette is built.

Fetch all accident occurrences from 2023:

crashes_2023 <- get_crashes(2023)
crashes_2023

You can request several years at once and filter by state (uf), federal highway (br), and — for accident data — severity:

# Fatal accidents in São Paulo and Rio de Janeiro, 2020-2023
fatal <- get_crashes(2020:2023, uf = c("SP", "RJ"), severity = "fatal")

# People involved in accidents on the BR-101
people_101 <- get_accidents(2023, br = 101)

# Traffic violations in Minas Gerais
violations_mg <- get_violations(2024, uf = "MG")

To see which years are available (and how large each file is) without downloading anything:

prf_years()

Understanding the variables

Each dataset has a bilingual codebook available offline. This runs without internet:

info_crashes()
#> # A tibble: 31 × 3
#>    variable       type  description             
#>    <chr>          <chr> <chr>                   
#>  1 id             int   Accident identifier     
#>  2 data_inversa   date  Accident date           
#>  3 dia_semana     chr   Day of week             
#>  4 horario        chr   Time of accident        
#>  5 uf             chr   State (UF)              
#>  6 br             int   Federal highway number  
#>  7 km             dbl   Highway kilometer marker
#>  8 municipio      chr   Municipality            
#>  9 causa_acidente chr   Accident cause          
#> 10 tipo_acidente  chr   Accident type           
#> # ℹ 21 more rows

Use lang = "pt" for descriptions in Portuguese:

head(info_violations(lang = "pt"))
#> # A tibble: 6 × 3
#>   variable                type  description                     
#>   <chr>                   <chr> <chr>                           
#> 1 numero_auto             chr   Número do auto de infração      
#> 2 dat_infracao            date  Data da infração                
#> 3 tip_abordagem           chr   Tipo de abordagem               
#> 4 ind_assinou_auto        chr   Indicador de assinatura do auto 
#> 5 ind_veiculo_estrangeiro chr   Indicador de veículo estrangeiro
#> 6 ind_sentido_trafego     chr   Sentido do tráfego

The full codebook is also available as a data object:

str(codebook)
#> tibble [97 × 5] (S3: tbl_df/tbl/data.frame)
#>  $ dataset       : chr [1:97] "accidents" "accidents" "accidents" "accidents" ...
#>  $ variable      : chr [1:97] "id" "pesid" "data_inversa" "dia_semana" ...
#>  $ type          : chr [1:97] "int" "int" "date" "chr" ...
#>  $ description_en: chr [1:97] "Record identifier" "Person identifier" "Accident date" "Day of week" ...
#>  $ description_pt: chr [1:97] "Identificador do registro" "Identificador da pessoa" "Data do acidente" "Dia da semana" ...

Managing the cache

prf_cache()
#> # A tibble: 0 × 4
#> # ℹ 4 variables: dataset <chr>, year <int>, size_mb <dbl>, cached_at <date>

prf_cache() lists the files currently cached; prf_cache_clear() removes them (all of them, or a specific dataset/year):

prf_cache_clear("violations", year = 2024)
prf_cache_clear()  # everything

Data sources

Raw data are published by the PRF on the federal government open data portal. The consolidation pipeline (raw CSV to Parquet, schema unification across years) is maintained at bonijoao/tidyprf-dados.