Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/timeplus-io/proton/llms.txt

Use this file to discover all available pages before exploring further.

Timeplus Proton provides a comprehensive set of SQL functions for stream processing, data transformation, and analysis. With over 600 built-in functions inherited from ClickHouse, plus streaming-specific functions, Proton offers powerful capabilities for real-time data processing.

​
Function Categories

​
Aggregate Functions

Aggregate functions perform calculations across multiple rows and return a single result. In streaming contexts, these functions support continuous aggregation over windows.
  • Basic aggregates: count(), sum(), avg(), min(), max()
  • Statistical: stddev(), variance(), corr(), covar()
  • Advanced: quantile(), median(), any(), group_array()
  • Specialized: arg_min(), arg_max(), uniq(), top_k()
See Aggregation Functions for details.

​
Window Functions

Window functions enable time-based windowing operations essential for stream processing.
  • Tumbling windows: tumble(stream, interval)
  • Hopping windows: hop(stream, slide, size)
  • Session windows: session(stream, timeout)
  • Window metadata: window_start, window_end
See Window Functions for details.

​
Date and Time Functions

Functions for working with timestamps, dates, and time zones.
  • Current time: now(), today(), now64()
  • Extraction: to_year(), to_month(), to_day_of_month(), to_hour(), to_minute(), to_second()
  • Conversion: to_date(), to_datetime(), to_unix_timestamp()
  • Arithmetic: add_years(), add_months(), subtract_days()
  • Formatting: format_datetime(), parse_datetime()
  • Time zones: Support for timezone-aware operations
See Time Functions for details.

​
String Functions

Functions for string manipulation and pattern matching.
  • Manipulation: concat(), substring(), upper(), lower(), trim()
  • Pattern matching: like, match(), extract()
  • Conversion: to_string(), cast()
  • Encoding: base64_encode(), base64_decode(), url_encode()

​
Array Functions

Functions for working with array data types.
  • Construction: array(), range(), array_concat()
  • Access: array_element(), array_slice()
  • Transformation: array_map(), array_filter(), array_reduce()
  • Aggregation: array_sum(), array_avg(), array_count()

​
Mathematical Functions

Standard mathematical operations and functions.
  • Arithmetic: plus, minus, multiply, divide, modulo
  • Rounding: round(), floor(), ceil(), trunc()
  • Exponential: exp(), log(), ln(), pow(), sqrt()
  • Trigonometric: sin(), cos(), tan(), asin(), acos(), atan()

​
Conditional Functions

Functions for conditional logic and branching.
  • Conditional: if(), multiIf(), case
  • Null handling: ifNull(), coalesce(), nullIf()
  • Type checking: isNull(), isNotNull()

​
Type Conversion

Functions for converting between data types.
  • Casting: CAST(x AS type), cast_or_default()
  • Conversion: to_int8(), to_float64(), to_string()

​
ClickHouse Function Compatibility

Timeplus Proton is built on ClickHouse and maintains compatibility with most ClickHouse functions. This means:
  • Extensive function library: Access to 600+ ClickHouse functions
  • Familiar syntax: If you know ClickHouse SQL, you know Proton
  • Documentation reference: ClickHouse function documentation generally applies
  • Performance: Optimized C++ implementations with SIMD acceleration

​
Streaming-Specific Enhancements

While maintaining ClickHouse compatibility, Proton adds streaming-specific capabilities:
  • Window functions: Native support for tumble, hop, and session windows
  • Streaming aggregation: Incremental computation with state management
  • Event time processing: Functions aware of event timestamps and watermarks
  • Stateful operations: Aggregations maintain state across streaming data

​
User-Defined Functions (UDF)

Timeplus Proton supports custom functions in Python and JavaScript, enabling you to extend the built-in function library with your own logic.

​
Python UDF

Create custom functions using Python: 
CREATE FUNCTION my_function(x INT) RETURNS INT
LANGUAGE PYTHON AS $$
def my_function(x):
    return x * 2
$$;

​
JavaScript UDF

Create custom functions using JavaScript: 
CREATE FUNCTION my_js_function(x INT) RETURNS INT
LANGUAGE JAVASCRIPT AS $$
function my_js_function(x) {
    return x * 2;
}
$$;

​
UDF Use Cases

  • Custom business logic: Implement domain-specific calculations
  • External API calls: Integrate with external services (with caution in streaming)
  • Complex transformations: Data enrichment and validation
  • Machine learning: Apply custom ML models to streaming data

​
Function Properties

​
Deterministic vs Non-Deterministic

  • Deterministic: Always return the same result for the same input (e.g., abs(), upper())
  • Non-deterministic: May return different results (e.g., now(), rand())

​
Null Handling

Most functions handle NULL values according to SQL standards:
  • Arithmetic operations with NULL return NULL
  • Use ifNull() or coalesce() to provide defaults
  • Some aggregates like count() have special NULL behavior

​
Case Sensitivity

Function names in Proton are case-insensitive:
  • COUNT(), count(), and Count() are equivalent
  • Improves compatibility with different SQL dialects

​
Performance Considerations

​
SIMD Optimization

Many functions use SIMD (Single Instruction, Multiple Data) for vectorized execution:
  • Processes multiple values simultaneously
  • Significant performance improvements for numeric operations
  • Automatic when supported by hardware

​
Streaming Context

In streaming queries, consider:
  • State size: Aggregations maintain state; use appropriate windows
  • Computation cost: Complex UDFs may impact latency
  • Memory usage: Large intermediate results can affect performance

​
Next Steps

Build docs developers (and LLMs) love

Get started for freeTalk to us