ClickHoused 0.1.2: native ClickHouse client for D (ddbc and HibernateD support coming)

[Vadim Lopatin]

On Thursday, 29 October 2026 at 21:26:21 UTC, Vadim Lopatin wrote:
> I'd like to announce **ClickHouseD**, a new client library for 
> [ClickHouse](https://clickhouse.com), the open-source 
> column-oriented database for analytics.

UPDATE: clickhoused 0.2.0 – 0.6.0: TLS, all column types, DDBC 
driver, performance

Since the 0.1.2 announcement, clickhoused, the native D client 
for ClickHouse (native TCP protocol, pure D, works with threads 
and vibe.d fibers), had five more releases. They are on 
code.dlang.org (https://code.dlang.org/packages/clickhoused); the 
full changelog is in the repository: 
https://github.com/buggins/clickhoused/blob/main/CHANGELOG.md

**0.2.0: TLS**
- Secure connections (port 9440) in thread and fiber mode: 
`secure=true` in the connection string or 
`ConnectionOptions.secure`, with CA file, client certificates 
(certificate-based login), SNI and `skip_verify` for development.
- OpenSSL 1.1.1 or 3.x is loaded at run time, only when the first 
secure connection is made: no extra dependency or build 
configuration for applications that don't use TLS. Trusted 
certificates come from the system store (also the Windows store 
on Windows).

**0.3.0: composite and extended types**
- `Array(T)`, `Tuple(...)` (named too), `Map(K, V)`, native 
`LowCardinality(T)`, `Decimal(P, S)` up to `Decimal256`, 
`Int128`/`UInt128`/`Int256`/`UInt256`, `IPv4`/`IPv6`.
- They map to D types in struct mapping and inserts: arrays to 
`T[]`, tuples to structs or `std.typecons.Tuple`, maps to 
associative arrays, plus value types `Decimal`, `Int128`…, 
`IPv4`, `IPv6`.

**0.4.0: more types**
- Geo types (`Point`, `Ring`, `LineString`, `Polygon`, 
`MultiPolygon`, ...) with a `GeoPoint` value type, `BFloat16`, 
`Time`/`Time64` (as `core.time.Duration`, or `TimeOfDay`).
- Fixed: `Float32`/`Float64` values as text now match the server 
exactly (the shortest decimal that reads back as the same value).

**0.5.0: Variant, Dynamic, JSON, Nested**
- `Variant(...)` (and `Geometry`), `Dynamic` and the new `JSON` 
type (as JSON text, with `std.json.JSONValue` mapping), and 
`Nested` with `flatten_nested = 0`.
- With this, every ClickHouse column type we planned is 
supported. `std.variant.Variant` maps to any column.

**0.6.0: performance**
- A new benchmark suite (`dub run :bench -b release`) found six 
hot spots; fixing them, without breaking changes, made SELECTs of 
typical columns about 4–5 times faster (about 6 times with LZ4 
compression), row-by-row inserts and `query!T` struct mapping 
about 3 times, and float-to-text conversion 10–20 times.
- Against a release ClickHouse 26.8 server, numeric results come 
in as fast as with ClickHouse's own `clickhouse-client`; results 
with strings take about 1.5–2 times its time. Numbers and method: 
https://github.com/buggins/clickhoused/blob/main/doc/benchmarks.md

**DDBC support**
DDBC (https://github.com/buggins/ddbc) now has a ClickHouse 
driver built on clickhoused: use the `ClickHouse` configuration 
and a URL such as 
`ddbc:clickhouse://localhost:9000/default?user=default`. It came 
with DDBC 0.8.0 and was updated with every clickhoused release 
since (DDBC 0.8.1–0.8.5; 0.8.2 also lets the PostgreSQL and ODBC 
drivers read many more column types). The versions are tagged and 
released on GitHub, but code.dlang.org does not import new DDBC 
versions automatically and still lists 0.7.0. Until it does, you 
can depend on the release commit directly (dub wants the commit 
hash here; this is v0.8.5):

```json
"dependencies": {
     "ddbc": { "repository": 
"git+https://github.com/buggins/ddbc.git", "version": 
"f584cb83c30818b693ab47cf1172ac51074f12de" }
},
"subConfigurations": { "ddbc": "ClickHouse" }
```

clickhoused is BSL-1.0, needs a D frontend 2.097 or newer (DMD 
2.097+, LDC 1.27+, GDC 12+), and is tested against ClickHouse 
23.8 to the latest release on Linux, Windows and macOS. It is an 
unofficial project, not affiliated with ClickHouse, Inc.

Next: an API review and freeze towards 1.0. Feedback, issues and 
pull requests are welcome: https://github.com/buggins/clickhoused