quickhouse 0.3.1__tar.gz → 0.3.3__tar.gz

{quickhouse-0.3.1 → quickhouse-0.3.3}/Cargo.lock

@@ -606,6 +606,7 @@ dependencies = [
  "android-tzdata",
  "iana-time-zone",
  "num-traits",
+ "serde",
  "windows-targets",
 ]
 
@@ -686,6 +687,16 @@ dependencies = [
  "tiny-keccak",
 ]
 
+[[package]]
+name = "core-foundation"
+version = "0.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b2a6cd9ae233e7f62ba4e9353e81a88df7fc8a5987b8d445b4d90c879bd156f6"
+dependencies = [
+ "core-foundation-sys",
+ "libc",
+]
+
 [[package]]
 name = "core-foundation-sys"
 version = "0.8.7"
@@ -1457,6 +1468,12 @@ version = "1.10.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6dbf3de79e51f3d586ab4cb9d5c3e2c14aa28ed23d180cf89b4df0454a69cc87"
 
+[[package]]
+name = "humantime"
+version = "2.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "15cdd26707701c53297e2fa6afb323d55fbc1d0810c3aec078ae3ef0424c3c15"
+
 [[package]]
 name = "hybrid-array"
 version = "0.4.13"
@@ -1497,6 +1514,7 @@ dependencies = [
  "hyper",
  "hyper-util",
  "rustls",
+ "rustls-native-certs",
  "tokio",
  "tokio-rustls",
  "tower-service",
@@ -1701,6 +1719,12 @@ dependencies = [
  "rustversion",
 ]
 
+[[package]]
+name = "integer-encoding"
+version = "3.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8bb03732005da905c88227371639bf1ad885cc712789c011c31c5fb3ab3ccf02"
+
 [[package]]
 name = "ipnet"
 version = "2.12.0"
@@ -1915,6 +1939,16 @@ dependencies = [
  "regex-automata",
 ]
 
+[[package]]
+name = "md-5"
+version = "0.10.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d89e7ee0cfbedfc4da3340218492196241d89eefb6dab27de5df917a6d2e78cf"
+dependencies = [
+ "cfg-if",
+ "digest 0.10.7",
+]
+
 [[package]]
 name = "md-5"
 version = "0.11.0"
@@ -2190,12 +2224,57 @@ dependencies = [
  "objc2-core-foundation",
 ]
 
+[[package]]
+name = "object_store"
+version = "0.11.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3cfccb68961a56facde1163f9319e0d15743352344e7808a11795fb99698dcaf"
+dependencies = [
+ "async-trait",
+ "base64 0.22.1",
+ "bytes",
+ "chrono",
+ "futures",
+ "humantime",
+ "hyper",
+ "itertools 0.13.0",
+ "md-5 0.10.6",
+ "parking_lot",
+ "percent-encoding",
+ "quick-xml",
+ "rand 0.8.7",
+ "reqwest",
+ "ring",
+ "serde",
+ "serde_json",
+ "snafu",
+ "tokio",
+ "tracing",
+ "url",
+ "walkdir",
+]
+
 [[package]]
 name = "once_cell"
 version = "1.21.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9f7c3e4beb33f85d45ae3e3a1792185706c8e16d043238c593331cc7cd313b50"
 
+[[package]]
+name = "openssl-probe"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7c87def4c32ab89d880effc9e097653c8da5d6ef28e6b539d313baaacfbafcbe"
+
+[[package]]
+name = "ordered-float"
+version = "2.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "68f19d67e5a2795c94e73e0bb1cc1a7edeb2e28efd39e2e1c9b7a40c1108b11c"
+dependencies = [
+ "num-traits",
+]
+
 [[package]]
 name = "parking_lot"
 version = "0.12.5"
@@ -2219,6 +2298,45 @@ dependencies = [
  "windows-link",
 ]
 
+[[package]]
+name = "parquet"
+version = "53.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2f8cf58b29782a7add991f655ff42929e31a7859f5319e53db9e39a714cb113c"
+dependencies = [
+ "ahash 0.8.12",
+ "arrow-array",
+ "arrow-buffer",
+ "arrow-cast",
+ "arrow-data",
+ "arrow-ipc",
+ "arrow-schema",
+ "arrow-select",
+ "base64 0.22.1",
+ "bytes",
+ "chrono",
+ "futures",
+ "half",
+ "hashbrown 0.15.5",
+ "num",
+ "num-bigint",
+ "object_store",
+ "paste",
+ "seq-macro",
+ "snap",
+ "thrift",
+ "tokio",
+ "twox-hash",
+ "zstd",
+ "zstd-sys",
+]
+
+[[package]]
+name = "paste"
+version = "1.0.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "57c0d7b74b563b49d38dae00a0c37d4d6de9b432382b2892f0574ddcae73fd0a"
+
 [[package]]
 name = "pem"
 version = "3.0.6"
@@ -2303,7 +2421,7 @@ dependencies = [
  "bytes",
  "fallible-iterator",
  "hmac",
- "md-5",
+ "md-5 0.11.0",
  "memchr",
  "rand 0.10.2",
  "sha2 0.11.0",
@@ -2504,9 +2622,19 @@ dependencies = [
  "syn 2.0.119",
 ]
 
+[[package]]
+name = "quick-xml"
+version = "0.37.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "331e97a1af0bf59823e6eadffe373d7b27f485be8748f71471c662c1f269b7fb"
+dependencies = [
+ "memchr",
+ "serde",
+]
+
 [[package]]
 name = "quickhouse-core"
-version = "0.3.1"
+version = "0.3.3"
 dependencies = [
  "anyhow",
  "arrow",
@@ -2521,6 +2649,8 @@ dependencies = [
  "google-cloud-gax",
  "google-cloud-googleapis",
  "mysql_async",
+ "object_store",
+ "parquet",
  "prost",
  "prost-types",
  "reqwest",
@@ -2540,7 +2670,7 @@ dependencies = [
 
 [[package]]
 name = "quickhouse-py"
-version = "0.3.1"
+version = "0.3.3"
 dependencies = [
  "pyo3",
  "quickhouse-core",
@@ -2744,6 +2874,7 @@ dependencies = [
  "encoding_rs",
  "futures-core",
  "futures-util",
+ "h2",
  "http",
  "http-body",
  "http-body-util",
@@ -2758,6 +2889,7 @@ dependencies = [
  "pin-project-lite",
  "quinn",
  "rustls",
+ "rustls-native-certs",
  "rustls-pki-types",
  "serde",
  "serde_json",
@@ -2883,6 +3015,18 @@ dependencies = [
  "zeroize",
 ]
 
+[[package]]
+name = "rustls-native-certs"
+version = "0.8.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dab5152771c58876a2146916e53e35057e1a4dfa2b9df0f0305b07f611fdea4d"
+dependencies = [
+ "openssl-probe",
+ "rustls-pki-types",
+ "schannel",
+ "security-framework",
+]
+
 [[package]]
 name = "rustls-pemfile"
 version = "2.2.0"
@@ -2894,9 +3038,9 @@ dependencies = [
 
 [[package]]
 name = "rustls-pki-types"
-version = "1.15.0"
+version = "1.15.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "764899a24af3980067ee14bc143654f297b22eaebfe3c7b6b211920a5a59b046"
+checksum = "2f4925028c7eb5d1fcdaf196971378ed9d2c1c4efc7dc5d011256f76c99c0a96"
 dependencies = [
  "web-time",
  "zeroize",
@@ -2926,12 +3070,30 @@ version = "1.0.23"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9774ba4a74de5f7b1c1451ed6cd5285a32eddb5cccb8cc655a4e50009e06477f"
 
+[[package]]
+name = "same-file"
+version = "1.0.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "93fc1dc3aaa9bfed95e02e6eadabb4baf7e3078b0bd1b4d7b6b0b68378900502"
+dependencies = [
+ "winapi-util",
+]
+
 [[package]]
 name = "saturating"
 version = "0.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ece8e78b2f38ec51c51f5d475df0a7187ba5111b2a28bdc761ee05b075d40a71"
 
+[[package]]
+name = "schannel"
+version = "0.1.29"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "91c1b7e4904c873ef0710c1f407dde2e6287de2bebc1bbbf7d430bb7cbffd939"
+dependencies = [
+ "windows-sys 0.61.2",
+]
+
 [[package]]
 name = "scopeguard"
 version = "1.2.0"
@@ -2944,12 +3106,41 @@ version = "4.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1c107b6f4780854c8b126e228ea8869f4d7b71260f962fefb57b996b8959ba6b"
 
+[[package]]
+name = "security-framework"
+version = "3.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b7f4bc775c73d9a02cde8bf7b2ec4c9d12743edf609006c7facc23998404cd1d"
+dependencies = [
+ "bitflags 2.13.1",
+ "core-foundation",
+ "core-foundation-sys",
+ "libc",
+ "security-framework-sys",
+]
+
+[[package]]
+name = "security-framework-sys"
+version = "2.17.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6ce2691df843ecc5d231c0b14ece2acc3efb62c0a398c7e1d875f3983ce020e3"
+dependencies = [
+ "core-foundation-sys",
+ "libc",
+]
+
 [[package]]
 name = "semver"
 version = "1.0.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8a7852d02fc848982e0c167ef163aaff9cd91dc640ba85e263cb1ce46fae51cd"
 
+[[package]]
+name = "seq-macro"
+version = "0.3.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1bc711410fbe7399f390ca1c3b60ad0f53f80e95c5eb935e52268a0e2cd49acc"
+
 [[package]]
 name = "serde"
 version = "1.0.229"
@@ -3101,6 +3292,33 @@ version = "1.15.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8ed6a63f02c8539c91a8685a86f4099661ba3da017932f6ebbea6de3f0fa7c90"
 
+[[package]]
+name = "snafu"
+version = "0.8.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6e84b3f4eacbf3a1ce05eac6763b4d629d60cbc94d632e4092c54ade71f1e1a2"
+dependencies = [
+ "snafu-derive",
+]
+
+[[package]]
+name = "snafu-derive"
+version = "0.8.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c1c97747dbf44bb1ca44a561ece23508e99cb592e862f22222dcf42f51d1e451"
+dependencies = [
+ "heck",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.119",
+]
+
+[[package]]
+name = "snap"
+version = "1.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "199905e6153d6405f9728fe44daace35f8f837bbf830bb6e85fbd5828709a886"
+
 [[package]]
 name = "socket2"
 version = "0.5.10"
@@ -3299,6 +3517,17 @@ dependencies = [
  "cfg-if",
 ]
 
+[[package]]
+name = "thrift"
+version = "0.17.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7e54bc85fc7faa8bc175c4bab5b92ba8d9a3ce893d0e9f42cc455c8ab16a9e09"
+dependencies = [
+ "byteorder",
+ "integer-encoding",
+ "ordered-float",
+]
+
 [[package]]
 name = "time"
 version = "0.3.54"
@@ -3803,6 +4032,16 @@ version = "0.9.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
 
+[[package]]
+name = "walkdir"
+version = "2.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "29790946404f91d9c5d06f9874efddea1dc06c5efe94541a7d6863108e3a5e4b"
+dependencies = [
+ "same-file",
+ "winapi-util",
+]
+
 [[package]]
 name = "want"
 version = "0.3.1"
@@ -4323,18 +4562,18 @@ dependencies = [
 
 [[package]]
 name = "zstd-safe"
-version = "7.2.4"
+version = "7.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8f49c4d5f0abb602a93fb8736af2a4f4dd9512e36f7f570d66e65ff867ed3b9d"
+checksum = "54a3ab4db68cea366acc5c897c7b4d4d1b8994a9cd6e6f841f8964566a419059"
 dependencies = [
  "zstd-sys",
 ]
 
 [[package]]
 name = "zstd-sys"
-version = "2.0.16+zstd.1.5.7"
+version = "2.0.13+zstd.1.5.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "91e19ebc2adc8f83e43039e79776e3fda8ca919132d68a1fed6a5faca2683748"
+checksum = "38ff0f21cfee8f97d94cef41359e0c89aa6113028ab0291aa8ca0038995a95aa"
 dependencies = [
  "cc",
  "pkg-config",

{quickhouse-0.3.1 → quickhouse-0.3.3}/Cargo.toml

@@ -3,7 +3,7 @@ resolver = "2"
 members = ["crates/quickhouse-core", "crates/quickhouse-py"]
 
 [workspace.package]
-version = "0.3.1"
+version = "0.3.3"
 edition = "2021"
 license = "MIT"
 authors = ["quickhouse contributors"]
@@ -50,6 +50,21 @@ base64 = "0.22"
 # verify the wire format. Pinned to the versions the bigquery crate resolves.
 prost = "0.13"
 prost-types = "0.13"
+# For the opt-in S3 data-lake archival path (ClickHouse destination only):
+# pinned to the exact arrow 53.x release line so `RecordBatch`es already
+# flowing through this crate's decoders pass straight into the Parquet
+# writer with zero conversion — a newer `parquet` would pull in a second,
+# incompatible `arrow-array` version (verified via `cargo tree`).
+parquet = { version = "53", default-features = false, features = ["arrow", "async", "object_store", "zstd", "snap"] }
+# Pinned to exactly the version `parquet` 53.4.0 itself depends on
+# (object_store = "0.11.0") — a newer object_store would resolve as a SECOND,
+# incompatible copy in the dependency graph (verified: this crate's own
+# `ObjectStore`/`Path` types wouldn't type-check against parquet's).
+# `aws` feature builds on reqwest/rustls (not native-tls), matching this
+# workspace's pure-rustls stack; it also pulls in `aws-lc-rs` as a rustls
+# crypto backend, but the process-wide `ring` install in sync.rs already
+# handles exactly this "another dependency's rustls backend" scenario.
+object_store = { version = "=0.11.2", default-features = false, features = ["aws"] }
 
 # Faster release builds for the shipped wheel.
 [profile.release]

{quickhouse-0.3.1 → quickhouse-0.3.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: quickhouse
-Version: 0.3.1
+Version: 0.3.3
 Classifier: Development Status :: 3 - Alpha
 Classifier: Intended Audience :: Developers
 Classifier: License :: OSI Approved :: MIT License
@@ -15,6 +15,8 @@ Requires-Dist: psycopg[binary]>=3.1 ; extra == 'test'
 Requires-Dist: pymysql>=1.1 ; extra == 'test'
 Requires-Dist: clickhouse-connect>=0.7 ; extra == 'test'
 Requires-Dist: tqdm>=4.60 ; extra == 'test'
+Requires-Dist: boto3>=1.34 ; extra == 'test'
+Requires-Dist: pyarrow>=14 ; extra == 'test'
 Provides-Extra: progress
 Provides-Extra: test
 License-File: LICENSE
@@ -131,6 +133,26 @@ Application Default Credentials. As a **destination** it also takes
 `write_method`: the default `"insert_all"` (simple, proven) or the opt-in
 `"storage_write"` (the gRPC Storage Write API — free and higher-throughput).
 
+A ClickHouse destination can also archive every synced batch to S3 as a data
+lake — a secondary, best-effort-free backup independent of ClickHouse's own
+retention:
+
+```python
+qh.ClickHouse(
+    "http://host:8123", database="analytics",
+    archive=qh.S3Archive(bucket="my-data-lake", prefix="quickhouse"),
+)
+```
+
+This streams Parquet — one file per parallel partition, never fully buffered
+in memory — to `s3://{bucket}/{prefix}/{dest_table}/dt=<date>/run=<id>/
+part-<partition>.parquet`, a Hive-style layout directly queryable by Athena,
+Spark, or DuckDB. Credentials fall back to the standard AWS chain (env vars,
+IAM role) unless overridden; pass `endpoint=` for an S3-compatible service
+like MinIO. A persistent upload failure fails the whole `sync()` call, same as
+a ClickHouse insert failure. Storage/request costs are billed by AWS as usual
+(free on a self-hosted MinIO).
+
 The DDL knobs (`engine`, `partition_by`, `order_by`, `primary_key`, `key`) are
 interpreted per destination — for ClickHouse they shape the `MergeTree`
 DDL; for BigQuery they map to partitioning and clustering. quickhouse creates
@@ -144,7 +166,13 @@ atomically — a crash mid-run never leaves the destination partial. For a
 BigQuery destination that swap runs as a query (a billed scan of the staged
 data), not a free copy job — BigQuery's copy jobs can silently skip rows still
 sitting in a table's streaming buffer, so a real query is what keeps this
-correct rather than just fast.
+correct rather than just fast. One accepted tradeoff on the ClickHouse path:
+an insert retried after a lost acknowledgment (not after a crash — the
+transfer is still running) can duplicate one batch's rows in the staging
+table, since `mode="full"` has no engine-level dedup like `ReplacingMergeTree`
+— rare, and harmless for `key`-based incremental syncs, but worth knowing if
+you see an unexpected small over-count on a full-refresh right after a
+transient network blip.
 
 **Incremental** tracks a high-water mark (the `watermark` column) in a small
 state table in the destination and copies only newer rows. Updated rows are
@@ -152,6 +180,12 @@ deduplicated on `key` — via ClickHouse's `ReplacingMergeTree`, or a `MERGE`
 upsert on BigQuery (where `key` is therefore required). Re-running with no new
 data does nothing.
 
+Both modes stage through a per-run-unique table (`{dest}_quickhouse_tmp_<id>`)
+that's dropped when the run finishes, including on failure. The unique name is
+what makes rapid re-runs and whole-call retries safe on BigQuery, whose
+streaming ingestion rejects writes into a table recently recreated under the
+same name.
+
 For daily syncs that need to catch late-arriving or edited rows, set
 `lookback_seconds` to re-scan a trailing window (e.g. `3 * 86400` for the last
 three days) — the dedup above keeps that overlap from creating duplicates.
@@ -196,8 +230,14 @@ timestamps across as-is, and booleans preserved. A few deliberate choices worth
 knowing:
 
 - **Arbitrary-precision decimals** (`numeric`/`DECIMAL`/`NUMERIC`) default to
-  `Float64`, since precision can't be recovered from the type alone. Pin an
-  exact type with `type_overrides` (e.g. `"Decimal(18, 2)"`).
+  `Float64`, since precision can't be recovered from the type alone — pin an
+  exact type with `type_overrides` (e.g. `"Decimal(18, 2)"`, `P <= 38`) and the
+  value is decoded exactly (no `Float64` round-trip), not just declared with
+  the right destination type. A value that doesn't fit the declared precision,
+  or is NaN/Infinity (PostgreSQL `numeric` only), coerces to `NULL` with a
+  warning, same as the out-of-range-date handling below. `P > 38`
+  (`Decimal256`) isn't supported yet and is rejected as a config error up
+  front, rather than silently falling back to `Float64`.
 - **`TIME`** columns transfer as canonical text into a `String` column
   (ClickHouse has no time-of-day type).
 - **Out-of-range dates** (and MySQL zero-dates like `0000-00-00`) coerce to

{quickhouse-0.3.1 → quickhouse-0.3.3}/README.md

@@ -101,6 +101,26 @@ Application Default Credentials. As a **destination** it also takes
 `write_method`: the default `"insert_all"` (simple, proven) or the opt-in
 `"storage_write"` (the gRPC Storage Write API — free and higher-throughput).
 
+A ClickHouse destination can also archive every synced batch to S3 as a data
+lake — a secondary, best-effort-free backup independent of ClickHouse's own
+retention:
+
+```python
+qh.ClickHouse(
+    "http://host:8123", database="analytics",
+    archive=qh.S3Archive(bucket="my-data-lake", prefix="quickhouse"),
+)
+```
+
+This streams Parquet — one file per parallel partition, never fully buffered
+in memory — to `s3://{bucket}/{prefix}/{dest_table}/dt=<date>/run=<id>/
+part-<partition>.parquet`, a Hive-style layout directly queryable by Athena,
+Spark, or DuckDB. Credentials fall back to the standard AWS chain (env vars,
+IAM role) unless overridden; pass `endpoint=` for an S3-compatible service
+like MinIO. A persistent upload failure fails the whole `sync()` call, same as
+a ClickHouse insert failure. Storage/request costs are billed by AWS as usual
+(free on a self-hosted MinIO).
+
 The DDL knobs (`engine`, `partition_by`, `order_by`, `primary_key`, `key`) are
 interpreted per destination — for ClickHouse they shape the `MergeTree`
 DDL; for BigQuery they map to partitioning and clustering. quickhouse creates
@@ -114,7 +134,13 @@ atomically — a crash mid-run never leaves the destination partial. For a
 BigQuery destination that swap runs as a query (a billed scan of the staged
 data), not a free copy job — BigQuery's copy jobs can silently skip rows still
 sitting in a table's streaming buffer, so a real query is what keeps this
-correct rather than just fast.
+correct rather than just fast. One accepted tradeoff on the ClickHouse path:
+an insert retried after a lost acknowledgment (not after a crash — the
+transfer is still running) can duplicate one batch's rows in the staging
+table, since `mode="full"` has no engine-level dedup like `ReplacingMergeTree`
+— rare, and harmless for `key`-based incremental syncs, but worth knowing if
+you see an unexpected small over-count on a full-refresh right after a
+transient network blip.
 
 **Incremental** tracks a high-water mark (the `watermark` column) in a small
 state table in the destination and copies only newer rows. Updated rows are
@@ -122,6 +148,12 @@ deduplicated on `key` — via ClickHouse's `ReplacingMergeTree`, or a `MERGE`
 upsert on BigQuery (where `key` is therefore required). Re-running with no new
 data does nothing.
 
+Both modes stage through a per-run-unique table (`{dest}_quickhouse_tmp_<id>`)
+that's dropped when the run finishes, including on failure. The unique name is
+what makes rapid re-runs and whole-call retries safe on BigQuery, whose
+streaming ingestion rejects writes into a table recently recreated under the
+same name.
+
 For daily syncs that need to catch late-arriving or edited rows, set
 `lookback_seconds` to re-scan a trailing window (e.g. `3 * 86400` for the last
 three days) — the dedup above keeps that overlap from creating duplicates.
@@ -166,8 +198,14 @@ timestamps across as-is, and booleans preserved. A few deliberate choices worth
 knowing:
 
 - **Arbitrary-precision decimals** (`numeric`/`DECIMAL`/`NUMERIC`) default to
-  `Float64`, since precision can't be recovered from the type alone. Pin an
-  exact type with `type_overrides` (e.g. `"Decimal(18, 2)"`).
+  `Float64`, since precision can't be recovered from the type alone — pin an
+  exact type with `type_overrides` (e.g. `"Decimal(18, 2)"`, `P <= 38`) and the
+  value is decoded exactly (no `Float64` round-trip), not just declared with
+  the right destination type. A value that doesn't fit the declared precision,
+  or is NaN/Infinity (PostgreSQL `numeric` only), coerces to `NULL` with a
+  warning, same as the out-of-range-date handling below. `P > 38`
+  (`Decimal256`) isn't supported yet and is rejected as a config error up
+  front, rather than silently falling back to `Float64`.
 - **`TIME`** columns transfer as canonical text into a `String` column
   (ClickHouse has no time-of-day type).
 - **Out-of-range dates** (and MySQL zero-dates like `0000-00-00`) coerce to

{quickhouse-0.3.1 → quickhouse-0.3.3}/crates/quickhouse-core/Cargo.toml

@@ -36,6 +36,8 @@ time.workspace = true
 base64.workspace = true
 prost.workspace = true
 prost-types.workspace = true
+parquet.workspace = true
+object_store.workspace = true
 
 [dev-dependencies]
 anyhow.workspace = true