GotRo is abbreviation of Gotong Royong
. the meaning in Indonesia
: "do it together", "mutual cooperation".
This Framework is rewrite of gokil, that previously use httprouter but rewritten using fasthttprouter (W
framework, deprecated), latest framework W2
are now using fiber.
versioning using this format 1.(M+(YEAR-2021)*12)DD
.HMM
,
so for example v1.213.1549 means it was released at 2021-02-13 15:49
- As similar as possible to gokil that still used by my old company (2014-now).
- Opinionated (choose the best dependency), for example by default uses
int64
andfloat64
for values, anduint64
for id(s). - 1-letter supporting package so we only need to write a short common function, such as:
I.ToS(1234)
to convertint64
tostring
)- A - Array
- B - Boolean
- C - Character (or Rune)
- D - Database
- F - Floating Point
- L - Logging
- M - Map
- I - Integer
- S - String
- T - Time (and Date)
- W - Web (the old "web framework") usable since 2017-03-08 DEPRECATED
- W2 - Web (the new codegen-based "web-framework") STATUS: usable since 2021-08-30, see W2/internal/example
- X - Anything (aka
any
) - Z - Z-Template Engine, that has syntax similar to ruby string interpolation
#{foo}
or any other that javascript friendly{/* foo */}
,[/* bar */]
,/*! bar */
- Comment and examples on each type and function, so it can be viewed using godoc, something like:
godoc github.com/kokizzu/gotro/A
Usable 3rd party database adapter:
- Ch = Clickhouse (OLAP, have migration tool) -- recommended
- Es = ElasticSearch (full text search, query only)
- Ms = Meilisearch (full text search)
- Pg = PostgreSQL (OLTP, using JSONB)
- Ql = QLDB (please use better database -_-)
- Rc = BigCache
- Rd = Redis works also for KvRocks, KeyDB, etc
- Tt = Tarantool (OLTP and cache, have migration tool) -- recommended
Other than above, you must use officially provided database adapter from respective vendors. For docker compose example. you can see local-docker-db
Benchmarked using hey -c 255 -n 255000 http://localhost:3001
on i7-4720HQ gotro almost 2x faster than gokil (12K
rps vs 23K
rps, thanks to fasthttp
),
this already includes session loading and template rendering (real-life use case, but with template auto-reloading which should be faster on production mode, since unlike in development mode it doesn't stat disk at all).
For newer framework W2
can achieve 20K
rps on Ryzen3 3100 without session loading and template rendering (only renders JSON), but with debug logging turned on (development mode), 33K
to 144K
rps on newer 32-core 128GB RAM NVMe server, see BENCHMARK.md for detailed result.
go get -u -v github.com/kokizzu/gotro
or for Go 1.16+: go mod download github.com/kokizzu/gotro
or just import one of the sub-library and run go mod tidy
- Ahmad Akbar Fauzi
- Anton Sofyan
- Dikaimin Simon
- Dimas Yudha Prawira
- Marcelinus Devin Yonas
- Michael Lim
- Muhammad Andri Juliansyah Putra
- Pham Hoang Tien
- Rizal Widyarta Gowandy
- add kardios/service for W2
- fix mysql adapter so it becomes usable (currently copied from Postgres'), probably wait until mysql has indexable json column, or do alters like scylladb and sqlite, or just remove and rewrite one for TiDB
- rewrite
D/Pg
using prepared statements, so no moreS.Z
- use
nikoksr/notify
for notification and mail sending instead of tied toW
- possibly refactor move cachedquery, records, etc to
D
package since nothing different about them - Review and benchmark which other databases we must support primarily for
D
, that can be silver bullet for extreme cases (high-write: sharding/partitioning and multi-master replication or auto-failover; full-text-search)- Aerospike <-- KV use case
- ActorDB <-- high-write
- CockroachDB <-- high-write (postgresql-compatible)
- Couchbase <-- high-write
- DGraph
- CrateDB <-- high-write
- GridDB <-- high-write
- GunDB
- IceFireDB <-- high-write (redis-compatible)
- LocustDB <-- OLAP use-cases
- NebulaGraph
- OrientDB
- PostgreXL <-- high-write (postgresql-compatible)
- SingleStore <-- high-write (mysql-compatible)
- TiDB <-- high-write (mysql-compatible)
- TypeSense
- YugaByteDB <-- high-write (postgresql/redis/cassandra-compatible)
- Also benchmark search engines (insert, force reindex duration, search substring first/last word foreach dataset, delete first 100 records by id)
- Add CDC example from TiDB to RedPanda to Materialize.io
- Pipe all request and respose W2/internal/example to Clickhouse, need to censor all the session key using S.XXH3
- Create metrics logger on W2/internal/example that push to redpanda then materialize.io
- Add ephemeral and/or persisted queuing/pub-sub service we're gonna use (NATS -- at most once delivery, RedPanda -- at least once delivery), see hugedbbench
- Add ExampleXxx function, getting started and more documentation
- Try other alternate graceful restart (zero downtime deployment): grace or endless instead of just overseer
- Add Catch NotFound (rewrite the
Response.Body
) if no route and static file found - add Generics support comes up (so we can embed the database connection dependencies inside the context without casting interface)
- Make sure all
D/*/*
docker-compose and dockertest works, volumes commented, have note on how to connect client and documentation URL, add from hugedbbench - Create NBIO codegen for websocket presentation/transport layer.