Examples

One pipeline, three backends

Join, summarize, and rank sales on phibble's DataFrame, on SQLite, and on Polars, with timings and a check that every backend returns the same rows.

Uses phibble phplyr

Inputs

Sales rows

Backend

DataFrame (in memory) SQLite (SqlTable) Polars (PolarsTable)

Products per region Running

phplyr functions call the methods of phibble's Table contract, so one pipeline runs on several backends. DataFrame evaluates it in PHP. SqlTable compiles col() expressions and helpers such as mean() to SQL, and collect() runs the query. PolarsTable runs the same expressions in Polars through the php_polars extension. This pipeline joins sales to stores, sums revenue per region and product, and keeps each region's top products.

$top = $sales |>
    filter(col('units')->gt(0)) |>
    left_join($stores, by: 'store') |>
    group_by('region', 'product') |>
    summarize(orders: n(), revenue: sum(col('units')->mul(col('price'))), avg_price: mean('price')) |>
    group_by('region') |>
    mutate(rank: min_rank(desc('revenue'))) |>
    filter(col('rank')->le(3)) |>
    arrange('region', 'rank') |>
    collect();

$top = phplyr($sales)
    ->filter(col('units')->gt(0))
    ->leftJoin($stores, by: 'store')
    ->groupBy('region', 'product')
    ->summarize(orders: n(), revenue: sum(col('units')->mul(col('price'))), avg_price: mean('price'))
    ->groupBy('region')
    ->mutate(rank: min_rank(desc('revenue')))
    ->filter(col('rank')->le(3))
    ->arrange('region', 'rank')
    ->collect()
    ->get();

Pipeline time (ms): 24.4
Rows in: 10000
Rows out: 15

Choose another backend to check its rows against DataFrame.

Revenue of the top products in each region

Result

region	product	orders	revenue	avg_price	rank
Central	p44	49	30,533.28	58.23	1
Central	p45	41	29,228.55	53.53	2
Central	p17	50	29,185.46	49.35	3
East	p49	54	31,208.42	47.07	1
East	p30	56	27,478.25	45.28	2
East	p15	39	24,994.01	56.73	3
North	p13	43	29,461.78	64.90	1
North	p43	43	28,182.83	52.89	2
North	p20	50	26,918.09	50.73	3
South	p16	48	28,577.31	56.62	1
South	p31	44	28,503.83	58.78	2
South	p29	48	28,301.87	53.28	3
West	p43	41	29,264.92	56.60	1
West	p18	51	28,001.14	56.37	2
West	p35	48	25,946.65	55.53	3

The timing covers the pipeline only, on the site's server. Building the tables is not timed. The performance page has benchmarks at a million rows, with dplyr for comparison.

Script: site/vignettes/backends.php in phpstats/phplyr@d319dc5.