Comparing the performances of various DBMS using dplyr for large-ish data

100%

Comparing the performances of various DBMS using dplyr for large-ish data

✓ Published0🌍 Public

Last edited Nov 20, 2017

Created on Nov 20, 2017

This example compares the execution times of data manipulation operations across different database management systems using the `dplyr` and `dbplyr` R packages. It benchmarks a `group_by` with `summarise` and a `left_join` with `mutate` on a 50-million-row tibble, running the same queries against in-memory data frames, SQLite databases (in-memory and file-based), and PostgreSQL. The code uses `system.time` to measure the elapsed user and system times for each operation, recording results in a benchmark summary table. The visualization highlights performance differences across database backends for typical data wrangling tasks.

AI-generated description

Similar vizzes

Loading thumbnail…

Takes as input an Rds file for clustering module (.Rmd code)

This example compares the read performance of RDS and Feather file formats for a single-cell clustering module object. It benchmarks loading the entire RDS object versus reading its four components (tsne, normalized expression, metadata, and marker list) as separate Feather files. The code uses the `feather`, `microbenchmark`, and `tibble` libraries, along with `readRDS()` and `write_feather()` functions, to measure mean execution times. The visualization shows autoplots of benchmark results, highlighting the time differences between the two storage approaches.

FFloWuenne

60% match

Loading thumbnail…

Gist d6bacd4ce0d5835b801ba028f9b1c81b

This example benchmarks seven JavaScript object-cloning techniques using the `chuhai` benchmarking library, with `lodash`, `Object.assign`, and JSON methods. It compares operations per second for deep and shallow copies of 1,000 parsed CSV records fetched from a remote dataset. The visualization highlights relative performance differences through horizontal bar lengths, with the fastest method labeled explicitly. The code uses `babyparse` for CSV parsing and `platform` to report the runtime environment, while the benchmark logic is orchestrated via `suite.cb` and `s.bench` calls. Results are output as Markdown text, not rendered graphically.

HHypercubed

57% match

Loading thumbnail…

Benchmarking leaflet.extras::addHeatmap vs leaflet.extras::addWebGLHeatmap

This example benchmarks the performance of two Leaflet heatmap rendering methods, `addHeatmap` and `addWebGLHeatmap`, across point datasets of increasing size from 10,000 to 2 million points. Using R’s `microbenchmark` package, it compares the time to load pages containing these visualizations in the browser, with results showing the standard heatmap consistently outperforms the WebGL version for all tested dataset sizes. The rendered output uses Leaflet and the `leaflet.extras` package, with the analysis displayed via R markdown and `ggplot2`.

RRCura

55% match

Loading thumbnail…

Gist c11433a697bc7660ed4cde37278fb580

This example demonstrates how to query a DuckDB database from R using the `dbplyr` and `duckdb` packages, showing the results of a simple aggregation of railway station data grouped by country. The code first creates an in-memory DuckDB connection, defines a view over a CSV file containing station information, and registers it as a tibble. It then performs a grouped count operation using dplyr verbs, which is lazily translated to SQL, and finally materializes the result locally with `collect()`. The output displays the number of stations per country, revealing a clear geographic distribution across European nations.

RRCura

55% match

Loading thumbnail…

Connexion à PostgreSQL + histogramme avec ggplot2

This example demonstrates how to create a histogram directly from a PostgreSQL database table using R. The code first establishes a database connection with the `DBI` and `RPostgreSQL` packages, then uses `dplyr` and `dbplyr` to query the table as a remote data frame. The `ggplot2` library renders the histogram via `geom_histogram`, mapping a specified column to the aesthetic. Finally, the connection is explicitly closed with `dbDisconnect`. The code illustrates a workflow for plotting distributed data without loading it entirely into local memory.

RRCura

52% match

Loading thumbnail…

Gist e92adbf79e3256a0d3e222f140a0433b

This example visualizes the relative performance of 16 JavaScript event-emitter libraries across two benchmark scenarios, “emit objects” and “emit one parameter,” using the raw output of the benchmark runner. The data, drawn from a gist by Hypercubed, is rendered as a grouped horizontal bar chart where each bar’s length corresponds to operations per second (ops/sec), with libraries sorted by speed and the fastest highlighted. The chart uses a simple SVG-based custom layout (likely with d3-scale for axes and d3-array for sorting) rather than a high-level charting library, directly mapping the numeric values to pixel widths. The visualization does not show error margins or statistical significance, only the raw ops/sec values and the benchmark’s “Fastest is” annotation for each test.

HHypercubed

52% match

Loading thumbnail…

Data Summary

This example loads a Pokémon dataset from a CSV and uses Datalib to generate a text-based summary, displaying the number of rows, columns, and file size, followed by per-column summaries, in a <pre> block on the page. It highlights how Datalib can quickly profile the shape of a dataset.

FFanYang715

52% match

Loading thumbnail…

globby - chuhai

This example compares the performance of two glob-matching libraries—the local `globby` fork and the upstream `globby` master—against `glob-stream`, using the `chuhai` and `ava` benchmarking suites. It shows raw operations per second for three pattern scenarios (negative globs and multiple positive globs) across async and sync APIs, with results from 10 runs each. The code directly invokes `globby.sync()`, `globby()`, and `glob-stream`'s `create()` to measure throughput, while the markdown output lists the fastest implementation per test. Data is generated from a temporary directory containing 500 files per subfolder, and the rendering is plain text from benchmark results.

HHypercubed

52% match