Changelog • ClickHouseHTTP

Version 1.1.2

CRAN release: 2026-10-02

Declared support for dbplyr’s 2nd edition interface (dbplyr_edition()), fixing dplyr::tbl() and other dbplyr-based verbs which previously failed with “uses dbplyr’s 1st edition interface, which is no longer supported” #7.
Added a sql_translation() method so as.character()/as.integer()/ as.numeric()/as.logical()/as.Date()/as.POSIXct(), ^, Sys.Date()/Sys.time() and sd()/var() translate to ClickHouse’s own conversion/aggregate functions (toString, toInt64, toFloat64, toUInt8, toDate, toDateTime, pow, today, now, stddevSamp, varSamp) instead of dbplyr’s ANSI defaults, which ClickHouse doesn’t recognize. (Implemented with Claude Code)
Added sql_table_analyze() (no-op, ClickHouse has no ANALYZE) and db_connection_describe() (shows the ClickHouse server/version instead of just the connection class name) dbplyr methods. (Implemented with Claude Code)
Added a testthat suite auditing dbplyr/ClickHouse SQL compatibility: offline SQL-rendering checks (no server needed) plus an opt-in live round-trip suite (set CLICKHOUSE_TEST_HOST to run it against a real server). (Implemented with Claude Code)

Fixed timezone-qualified DateTime columns being reported as unsupported with format = "TabSeparatedWithNamesAndTypes".
format = "TabSeparatedWithNamesAndTypes" now uses the timezone embedded in a column’s type (e.g. DateTime('Europe/Paris')) when present, falling back to the session_timezone connection setting otherwise.

Arrow timestamp conversion now uses the session_timezone connection setting #6.
Tab-separated DateTime values now match Arrow timestamps when session_timezone is not defined.

CRAN release: 2026-05-18

Implemented dbColumnInfo() for ClickHouseHTTPResult: returns a data frame with name and type columns describing the R types of a query result. Works for both Arrow and TabSeparatedWithNamesAndTypes formats by parsing the buffered response content without consuming the result cursor, so dbFetch() can still be called afterwards.

Migrated HTTP backend from httr to httr2: all requests are now built using the httr2 request pipeline (request() → req_headers() → req_options() → req_perform()).
The reset_handle parameter of dbConnect() now maps to the httr2/curl fresh_connect option instead of the former httr::handle_reset(). Its behaviour is unchanged: set to TRUE to force a new TCP connection for each request rather than reusing a pooled one.
HTTP error responses are handled manually (via .query_success()) rather than relying on httr2’s automatic error raising, preserving existing ClickHouse exception-code detection logic.
Minimum R version bumped to 4.1 (required by the native pipe |> used in the httr2 request pipeline).

Bug fix (identified and corrected by Claude): Date columns returned as integers when using format="Arrow". Newer ClickHouse versions send Date as Arrow date32 directly, but the internal type-casting logic contained a Date32 → int32 rule that was only ever intended as an intermediate step for old ClickHouse (which sent Date as UInt16). When applied to an already correctly typed date32 column, it stripped the date semantics before conversion to R, yielding an integer instead of a Date. Fixed by rewriting the Arrow schema cast helpers (result.R) to generate the two-step UInt16 → int32 → date32 chain only when the source column is actually UInt16, leaving date32 and timestamp columns untouched.

CRAN release: 2025-06-05

CRAN release: 2024-04-18

CRAN release: 2023-07-04

Add the possibility to use httr::handle_reset() when calling httr::POST() to allow several independent connections in the same session: when needed, set the “reset_handle” parameter to TRUE when calling dbConnect()

CRAN release: 2023-02-16

path parameter in dbConnect() => use of a reverse proxy

CRAN release: 2022-09-05

quote="" when calling data.table::fread()

Explicit references to functions from dependencies

CRAN release: 2022-03-29

Better support of “TabSeparatedWithNamesAndTypes” format