Aggregator is a batch processing library for Go supports returning values. You can group up and process batch of tasks with keys in a single callback. Using it for grouping up database query or cache can help you to reduce loading of database and network.
This project may contain bugs and have not being tested at all. Use under your own risk, but feel free to test, make pull request and improve this project.
- Support multi Aggregators (using
AggregatorList) for fallback. - Support multi workers to flush tasks.
- Support Go generics for query keys and result values.
- Support timeout-only or tasks limit-only.
- Support singleflight (using singleflight-any).
Currently Go 1.18+ is required (for go generics), backward compatible is planned.
go get github.com/serkodev/aggregator@latestcallback := func(keys []string) (map[string]Book, error) {
results := db.Query(`SELECT * FROM books WHERE name IN ?`, keys)
return results, nil
}
agg, _ := aggregator.New(callback, 100*time.Millisecond, 5).Run()
for _, name := range []string{"foo", "foo", "bar", "baz", "baz"} {
go func(n string) {
book, err := agg.Query(n).Get()
if err == nil {
print(book.Name + ":" + book.Price, " ")
}
}(name)
}
// foo:10 foo:10 bar:25 baz:30 baz:30flowchart LR;
subgraph A [Aggregator]
direction TB
subgraph cb ["Customize Process (example)"]
direction TB
input("Input
[]string{#quot;foo#quot;, #quot;bar#quot;, #quot;baz#quot;}")
db[("Database
SELECT price FROM books<br />WHERE name
IN ('foo', 'bar', 'baz')")]
output("return map[string]int{
#quot;foo#quot;: 10,
#quot;bar#quot;: 25,
#quot;baz#quot;: 30,
}")
input --> db --> output
style output text-align:left
end
Wait -- Reach tasks limit / Timeout -->
cb --> rt("Return value to each Request")
end
req1[Request 1] --> q_foo_("Query(#quot;foo#quot;)"):::bgFoo --> A
req2[Request 2] --> q_foo2("Query(#quot;foo#quot;)"):::bgFoo --> A
req3[Request 3] --> q_bar_("Query(#quot;bar#quot;)"):::bgBar --> A
req4[Request 4] --> q_baz_("Query(#quot;baz#quot;)"):::bgBaz --> A
req5[Request 5] --> q_baz2("Query(#quot;baz#quot;)"):::bgBaz --> A
A --- rtn1("return 10"):::bgFoo --> req1_[Request 1]
A --- rtn2("return 10"):::bgFoo --> req2_[Request 2]
A --- rtn3("return 25"):::bgBar --> req3_[Request 3]
A --- rtn4("return 30"):::bgBaz --> req4_[Request 4]
A --- rtn5("return 30"):::bgBaz --> req5_[Request 5]
classDef bgFoo fill:green;
classDef bgBar fill:blue;
classDef bgBaz fill:purple;
AggregatorList contains a slice of Aggregator, you can create it by aggregator.NewList(...). If the prior order aggregator cannot return data for any keys. Then AggregatorList will query data from the next aggregator for fallback.
For example, you create an AggregatorList with cache and database aggregator, when the data has not been cached, it will auto query from database.
cacheAgg := aggregator.New(func(k []string) (map[string]string, error) {
fmt.Println("fetch from cache...", k)
return map[string]string{
"key1": "val1",
"key2": "val2",
}, nil
}, 50*time.Millisecond, 10)
databaseAgg := aggregator.New(func(k []string) (map[string]string, error) {
fmt.Println("fetch from database...", k)
return map[string]string{
"key1": "val1",
"key2": "val2",
"key3": "val3",
"key4": "val4",
}, nil
}, 50*time.Millisecond, 10)
list := aggregator.NewList(cacheAgg, databaseAgg).Run()
results := list.QueryMulti([]string{"key1", "key2", "key3", "key4"})
// fetch from cache... ["key1", "key2", "key3", "key4"]
// fetch from database... ["key3", "key4"]
// results: ["val1", "val2", "val3", "val4"]In some use case you may need to prevent cache breakdown. Aggregator works with singleflight by using singleflight-any (supports Go generics).
MIT License