parquet export docs

documentation/reference/api/rest.md

@@ -591,15 +591,42 @@ returned in a tabular form to be saved and reused as opposed to JSON.
 `/exp` is expecting an HTTP GET request with following parameters:
 
 | Parameter | Required | Description                                                                                                                                                                                                                  |
-| :-------- | :------- | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+|:----------|:---------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
 | `query`   | Yes      | URL encoded query text. It can be multi-line.                                                                                                                                                                                |
 | `limit`   | No       | Paging opp parameter. For example, `limit=10,20` will return row numbers 10 through to 20 inclusive and `limit=20` will return first 20 rows, which is equivalent to `limit=0,20`. `limit=-20` will return the last 20 rows. |
 | `nm`      | No       | `true` or `false`. Skips the metadata section of the response when set to `true`.                                                                                                                                            |
+| `fmt`     | No       | Export format. Valid values: `parquet`, `csv`. When set to `parquet`, exports data in Parquet format instead of CSV.                                                                                                         |
+
+#### Parquet Export Parameters
+
+:::warning
+
+Parquet exports currently require writing interim data to disk, and therefore must be run on **read-write instances only**.
+
+This limitation will be removed in future.
+
+:::
+
+When `fmt=parquet`, the following additional parameters are supported:
+
+| Parameter            | Required | Default   | Description                                                                                         |
+|:---------------------|:---------|:----------|:----------------------------------------------------------------------------------------------------|
+| `partition_by`       | No       | `NONE`    | Partition unit: `NONE`, `HOUR`, `DAY`, `WEEK`, `MONTH`, or `YEAR`.                                  |
+| `compression_codec`  | No       | `ZSTD`    | Compression algorithm: `UNCOMPRESSED`, `SNAPPY`, `GZIP`, `LZ4`, `ZSTD`, `LZ4_RAW`, `BROTLI`, `LZO`. |
+| `compression_level`  | No       | `9`       | Compression level (codec-specific). Higher values = better compression but slower.                  |
+| `row_group_size`     | No       | `100000`  | Number of rows per Parquet row group.                                                               |
+| `data_page_size`     | No       | `1048576` | Size of data pages in bytes (default 1MB).                                                          |
+| `statistics_enabled` | No       | `true`    | Enable Parquet column statistics: `true` or `false`.                                                |
+| `parquet_version`    | No       | `2`       | Parquet format version: `1` (v1.0) or `2` (v2.0).                                                   |
+| `raw_array_encoding` | No       | `true`    | Use raw encoding for arrays: `true` or `false`.                                                     |
+| `rmode`              | No       | `false`   | Set HTTP response mode: `nodelay` or not sent                                                       |
 
 The parameters must be URL encoded.
 
 ### Examples
 
+#### CSV Export (default)
+
 Considering the query:
 
 ```shell
@@ -620,6 +647,44 @@ A HTTP status code of `200` is returned with the following response body:
 200501BS00005,"2005-01-10T00:00:00.000Z",21:13
 ```
 
+#### Parquet Export
+
+Export query results to Parquet format:
+
+```shell
+curl -G \
+  --data-urlencode "query=SELECT * FROM trades WHERE timestamp IN today()" \
+  --data-urlencode "fmt=parquet" \
+  http://localhost:9000/exp > trades_today.parquet
+```
+
+#### Parquet Export with Custom Options
+
+Export with custom compression and partitioning:
+
+```shell
+curl -G \
+  --data-urlencode "query=SELECT * FROM trades" \
+  --data-urlencode "fmt=parquet" \
+  --data-urlencode "partition_by=DAY" \
+  --data-urlencode "compression_codec=ZSTD" \
+  --data-urlencode "compression_level=9" \
+  --data-urlencode "row_group_size=1000000" \
+  http://localhost:9000/exp > trades.parquet
+```
+
+#### Parquet Export with LZ4 Compression
+
+Export with LZ4_RAW compression for faster export:
+
+```shell
+curl -G \
+  --data-urlencode "query=SELECT symbol, price, amount FROM trades WHERE timestamp > dateadd('h', -1, now())" \
+  --data-urlencode "fmt=parquet" \
+  --data-urlencode "compression_codec=LZ4_RAW" \
+  http://localhost:9000/exp > recent_trades.parquet
+```
+
 ## Error responses
 
 ### Malformed queries

documentation/reference/function/aggregation.md

@@ -22,17 +22,19 @@ Running it will result in the following error:
 
 You can work around this limitation by using CTEs or subqueries:
 
-```questdb-sql title="aggregates as function args workaround" demo
+```questdb-sql title="CTE workaround"
 -- CTE
 WITH minmax AS (
-    SELECT min(timestamp) as min_date, max(timestamp) as max_date FROM trades
+  SELECT min(timestamp) AS min_date, max(timestamp) AS max_date FROM trades
 )
 SELECT datediff('d', min_date, max_date) FROM minmax;
 
 -- Subquery
-SELECT datediff('d', min_date, max_date) FROM (
-    SELECT min(timestamp) as min_date, max(timestamp) as max_date FROM trades
+SELECT datediff('d', min_date, max_date)
+FROM (
+  SELECT min(timestamp) AS min_date, max(timestamp) AS max_date FROM trades
 );
+
 ```
 
 :::
@@ -921,6 +923,49 @@ FROM (SELECT rnd_double() a FROM long_sequence(100));
 | :--------------- |
 | 49.5442334742831 |
 
+## mode
+
+`mode(value)` - calculates the mode (most frequent) value out of a particular dataset. 
+
+For `mode(B)`, if there are an equal number of `true` and `false` values, `true` will be returned as a tie-breaker.
+
+For other modes, if there are equal mode values, the returned value will be whichever the code identifies first.
+
+To make the result deterministic, you must enforce an underlying sort order.
+
+#### Parameters
+
+- `value` - one of (LONG, DOUBLE, BOOLEAN, STRING, VARCHAR, SYMBOL)
+
+#### Return value
+
+Return value type is the same as the type of the input `value`.
+
+
+#### Examples
+
+With this dataset:
+
+| symbol    | value |
+|-----------|-------|
+| A         | alpha |
+| A         | alpha |
+| A         | alpha |
+| A         | omega |
+| B         | beta  |
+| B         | beta  |
+| B         | gamma |
+
+```questdb-sql
+SELECT symbol, mode(value) as mode FROM dataset;
+```
+
+| symbol | mode  |
+|--------|-------|
+| A      | alpha |
+| B      | beta  |
+
+
 ## stddev / stddev_samp
 
 `stddev_samp(value)` - Calculates the sample standard deviation of a set of

documentation/reference/function/meta.md

@@ -589,6 +589,55 @@ If you want to re-read metadata for all user tables, simply use an asterisk:
 SELECT hydrate_table_metadata('*');
 ```
 
+## sys.copy_export_log
+
+`sys.copy_export_log` is a pseudo-table containing the export log for `COPY TO` operations.
+
+**Arguments:**
+
+- `sys.copy_export_log` does not require arguments.
+
+**Return value:**
+
+Returns metadata on `COPY TO` export operations for the last three days, including the columns:
+
+- `ts` - timestamp of the log event
+- `id` - export identifier that can be used to track export progress
+- `table_name` - source table name (or 'query' for subquery exports)
+- `export_path` - destination directory path for the export
+- `num_exported_files` - how many output files were written
+- `phase` - progress markers for each export step
+- `status` - event status for each phase, for example 'started', 'finished'
+- `message` - additional text (important for error rows)
+- `errors` - error number or flag
+
+**Examples:**
+
+```questdb-sql
+COPY trades TO 'trades' WITH FORMAT PARQUET;
+```
+
+| id               |
+|------------------| 
+| 38b2b45f28aa822e |
+
+Checking the log:
+
+```questdb-sql
+SELECT * FROM copy_export_log() WHERE id = '38b2b45f28aa822e';
+```
+
+| ts                          | id               | table_name | export_path                    | num_exported_files | phase                 | status   | message | errors |
+|-----------------------------|------------------|------------|--------------------------------|--------------------|-----------------------|----------|---------|--------|
+| 2025-10-27T14:07:20.513119Z | 38b2b45f28aa822e | trades     | null                           | null               | wait_to_run           | started  | queued  | 0      |
+| 2025-10-27T14:07:20.541779Z | 38b2b45f28aa822e | trades     | null                           | null               | wait_to_run           | finished | 0       |
+| 2025-10-27T14:07:20.542552Z | 38b2b45f28aa822e | trades     | null                           | null               | converting_partitions | started  | null    | 0      |
+| 2025-10-27T14:07:20.658111Z | 38b2b45f28aa822e | trades     | null                           | null               | converting_partitions | finished | null    | 0      |
+| 2025-10-27T14:07:20.658185Z | 38b2b45f28aa822e | trades     | null                           | null               | move_files            | started  | null    | 0      |
+| 2025-10-27T14:07:20.670200Z | 38b2b45f28aa822e | trades     | null                           | null               | move_files            | finished | null    | 0      |
+| 2025-10-27T14:07:20.670414Z | 38b2b45f28aa822e | trades     | /<dbroot>/export/trades/ | 26                 | success               | finished | null    | 0      |
+
+
 ## flush_query_cache()
 
 `flush_query_cache' invalidates cached query execution plans.