Skip to main content

​
Overview

TeeBI is designed for ultra-fast data processing through its array-based architecture. Every column or field is stored as a simple contiguous array, providing optimal memory locality and cache performance.

​
Array-Based Architecture

​
Column Storage

TeeBI stores data in columnar format, where each column is represented as a native array: 
TInt32Array = Array of Integer;
TInt64Array = Array of Int64;
TSingleArray = Array of Single;
TDoubleArray = Array of Double;
TTextArray = Array of String;
TDateTimeArray = Array of TDateTime;
TBooleanArray = Array of Boolean;

​
Performance Benefits

Memory Locality: Contiguous array storage means better CPU cache utilization and faster memory access patterns. Pointer Math: Enabled with {$POINTERMATH ON} for direct pointer arithmetic on arrays, avoiding overhead. Precision Control: Uses {$EXCESSPRECISION OFF} on x64 for faster floating-point operations with acceptable precision. Native Types: Uses TInteger = NativeInt to automatically optimize for 32-bit or 64-bit architecture.

​
Hybrid Sorting Algorithms

TeeBI uses optimized sorting algorithms combining QuickSort and InsertionSort:
  • QuickSort for large segments
  • InsertionSort for small segments where it outperforms QuickSort
  • Automatic algorithm selection based on data size

​
Speed Benchmarks

See the Speed demo for performance comparisons with other data structures.

​
Key Performance Metrics

  • In-Memory Operations: Direct array access without abstraction overhead
  • No Dependencies: Pure Pascal RTL implementation, no database server required
  • Zero-Copy: Many operations work directly on array pointers

​
Best Practices

​
Pre-allocate Memory

When building large datasets, pre-allocate array sizes to avoid repeated reallocations: 
Data.Resize(ExpectedRowCount);

​
Use Appropriate Data Types

Choose the smallest data type that fits your needs:
  • Use Int32 instead of Int64 when values fit in 32 bits
  • Use Single instead of Double for reduced precision requirements

​
Leverage Parallel Processing

For large datasets, use parallel operations (see Parallel Processing).

​
Enable Compression

When persisting or transmitting data, use compression to reduce I/O time (see Compression).

​
Performance Monitoring

Use Delphi’s profiling tools to identify bottlenecks:
  • Sampling Profiler: Available in RAD Studio
  • Stopwatch: Use TStopwatch from System.Diagnostics
uses System.Diagnostics;

var SW: TStopwatch;
SW := TStopwatch.StartNew;

// Your TeeBI operations here

SW.Stop;
ShowMessage('Elapsed: ' + SW.ElapsedMilliseconds.ToString + ' ms');

Build docs developers (and LLMs) love

Get started for freeTalk to us