Array Library

The array library offers easy allocation of large memory mapped files with less performance overhead than the traditional buffers[i].get(j)-style constructions java often leads to given its suffocating 2 Gb ByteBuffer size limitation.

It accomplishes this by delegating block oerations down to the appropriate page. If the operation crosses a page boundary, it is not delegated and a bit slower.

It's a very C++-style library that does unidiomatic things with interface default functions to get diamond inheritance.

Quick demo:

var array = LongArray.mmapForWriting(Path.of("/tmp/test"), 1<<16);

array.transformEach(50, 1000, (pos, val) -> Long.hashCode(pos));
array.quickSort(50, 1000);
if (array.binarySearch(array.get(100), 50, 1000) >= 0) {
    System.out.println("Nevermind, I found it!");
}

array.range(50, 1000).fill(0, 950, 1);
array.forEach(0, 100, (pos, val) -> {
    System.out.println(pos + ":" + val);
});

Query Buffers

The classes IntQueryBuffer and LongQueryBuffer are used heavily in the search engine's query processing.

They are dual-pointer buffers that offer tools for filtering data.

LongQueryBuffer buffer = new LongQueryBuffer(1000);

// later ...

// Prepare the buffer for filling
buffer.reset();
fillBufferSomehow(buffer); 

// length is updated and data is set
// read pointer and write pointer is now at 0

// A typical filtering operation may look like this:
        
while (buffer.hasMore()) { // read < end
    if (someCondition(buffer.currentValue())) {
        // copy the value pointed to by the read
        // pointer to the read pointer, and
        // advance both
        buffer.retainAndAdvance();
    }
    else {
        // advance the read pointer
        buffer.rejectAndAdvance();
    }
}

// set end to the write pointer, and 
// resets the read and write pointers
buffer.finalizeFiltering();

// ... after this we can filter again, or
// consume the data

Especially noteworthy are the operations retain() and reject() in IntArraySearch and LongArraySearch. They keep or remove all items in the buffer that exist in the referenced range of the array, which must be sorted.

These are used to offer an intersection operation for the B-Tree with sub-linear run time.

2.5 KiB Raw Blame History

Array Library

Quick demo:

Query Buffers

2.5 KiB

Raw Blame History