Outgrowing your laptop with Positron

useR! 2025 | Julia Silge

Hello!

Where do we start?

library(tidyverse)

df <- read_csv("pokemon_df.csv")
df |>
  group_by(type_1) |>
  summarize(hp = mean(hp)) |>
  arrange(-hp)
#> # A tibble: 18 × 2
#>    type_1      hp
#>    <chr>    <dbl>
#>  1 dragon    84.5
#>  2 normal    77.3
#>  3 fairy     72.9
#>  4 psychic   72.6
#>  5 fighting  71.6
#>  6 ground    71.6
#>  7 water     71.1
#>  8 flying    70.8
#>  9 ice       70.5
#> 10 fire      69.7
#> 11 dark      69.7
#> 12 steel     67.3
#> 13 poison    67.2
#> 14 grass     66.7
#> 15 rock      65.3
#> 16 ghost     64.2
#> 17 bug       57.7
#> 18 electric  55.8

Your laptop is a great place to start!

Memory
Performance

Lazy evaluation with duckdb

library(tidyverse)
library(duckplyr)

df <- read_parquet_duckdb("flights.parquet")
df |>
  summarize(.by = month, arr_delay = mean(arr_delay, na.rm = TRUE)) |>
  arrange(month)
#> # A duckplyr data frame: 2 variables
#>    month arr_delay
#>    <int>     <dbl>
#>  1     1     6.13 
#>  2     2     5.61 
#>  3     3     5.81 
#>  4     4    11.2  
#>  5     5     3.52 
#>  6     6    16.5  
#>  7     7    16.7  
#>  8     8     6.04 
#>  9     9    -4.02 
#> 10    10    -0.167
#> 11    11     0.461
#> 12    12    14.9

Lazy evaluation with duckdb

library(tidyverse)
library(duckplyr)

df <- read_parquet_duckdb("flights.parquet")
df |>
  summarize(.by = month, arr_delay = mean(arr_delay, na.rm = TRUE)) |>
  arrange(month) |>
  collect()
#> # A tibble: 12 × 2
#>    month arr_delay
#>    <int>     <dbl>
#>  1     1     6.13 
#>  2     2     5.61 
#>  3     3     5.81 
#>  4     4    11.2  
#>  5     5     3.52 
#>  6     6    16.5  
#>  7     7    16.7  
#>  8     8     6.04 
#>  9     9    -4.02 
#> 10    10    -0.167
#> 11    11     0.461
#> 12    12    14.9

Lazy evaluation with duckdb

library(tidyverse)
library(duckplyr)

urls <- glue::glue(
  "https://blobs.duckdb.org/flight-data-partitioned/Year={2022:2024}/data_0.parquet"
)

df <- read_parquet_duckdb(urls)
df |>
  summarize(.by = Month, ArrDelay = mean(ArrDelay, na.rm = TRUE)) |>
  arrange(Month)
#> # A duckplyr data frame: 2 variables
#>    Month ArrDelay
#>    <dbl>    <dbl>
#>  1     1     7.42
#>  2     2     3.08
#>  3     3     7.55
#>  4     4     7.64
#>  5     5     8.51
#>  6     6    12.3 
#>  7     7    13.3 
#>  8     8     8.58
#>  9     9     3.92
#> 10    10     1.62
#> 11    11     1.66
#> 12    12     6.89

Lazy evaluation with duckdb

Tips

Keep in mind that DuckDB is an engine for executing queries and Parquet is a file format
Working with lazy queries can take some getting used to
Translation between dplyr and DuckDB is amazing, but not everything is supported
If you need to read or write parquet files yourself, try out the nanoparquet package
Once files are not on your laptop, often the hardest part is… auth 😩

Connect to databases

library(tidyverse)

con <- connections::connection_open(RSQLite::SQLite(), "nycflights13.sqlite")
df <- tbl(con, "flights")
df |>
  group_by(month) |>
  summarise(arr_delay = mean(arr_delay, na.rm = TRUE))
#> # Source:   SQL [?? x 2]
#> # Database: sqlite 3.50.1 [/Users/juliasilge/Work/talks/useR-2025/nycflights13.sqlite]
#>    month arr_delay
#>    <int>     <dbl>
#>  1     1     6.13 
#>  2     2     5.61 
#>  3     3     5.81 
#>  4     4    11.2  
#>  5     5     3.52 
#>  6     6    16.5  
#>  7     7    16.7  
#>  8     8     6.04 
#>  9     9    -4.02 
#> 10    10    -0.167
#> 11    11     0.461
#> 12    12    14.9

Connect to databases

library(tidyverse)

con <- connections::connection_open(RSQLite::SQLite(), "nycflights13.sqlite")
df <- tbl(con, "flights")
df |>
  group_by(month) |>
  summarise(arr_delay = mean(arr_delay, na.rm = TRUE)) |>
  collect()
#> # A tibble: 12 × 2
#>    month arr_delay
#>    <int>     <dbl>
#>  1     1     6.13 
#>  2     2     5.61 
#>  3     3     5.81 
#>  4     4    11.2  
#>  5     5     3.52 
#>  6     6    16.5  
#>  7     7    16.7  
#>  8     8     6.04 
#>  9     9    -4.02 
#> 10    10    -0.167
#> 11    11     0.461
#> 12    12    14.9

Connect to databases

Tips

Keep in mind that it’s worth learning a little SQL, even if you use dbplyr
Once again, working with lazy queries can take some getting used to
Also once again, translation between dplyr and SQL is amazing, but subtleties sometime come up
You can also use the Connections Pane with DuckDB!
When working with databases, often the hardest part is… auth 😩

Remote SSH sessions

Tips

Keep in mind that your terminals, files, extensions, etc are on the remote server
You can control how long your sessions run via the kernelSupervisor.shutdownTimeout setting
You can also use these types of remote sessions all locally with Docker for reproducibility; check out Andrew Heiss’ recent blog post
When working with remote sessions, often the hardest part is… auth 😩

Want to try out Positron?

Visit https://positron.posit.co/ for documentation and installers
Find us on GitHub at https://github.com/posit-dev/positron
You can use Positron in a server environment on Posit Workbench

Thank you!

@juliasilge

@juliasilge.com

@juliasilge@fosstodon.org

youtube.com/juliasilge

juliasilge.com