Skip to content

wowczarek/rbt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

39 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

rbt

About

Yet another red-black tree implementation, written in C99. Because I needed one for a dictionary / general-purpose dynamic index (to be used for Barser). Three-pointer (parent + two-child array), plus value pointer and uint32_t keys, node colour as extra bool. No pointer bit reuse, nothing too clever. No thread safety or cache awareness. As basic as it gets, no-nonsense code. Completely non-recursive, although using stacks and FIFO queues. This implementation can alternatively be referred to as Random Bastard Tree. The code is BSD 2-clause licenced. Why not GPL? For because no, forced freedom is not freedom in my book.

Supports:

  • retrieval, insertion, deletion (bottom-up),
  • verification of red-black tree invariants / correctness,
  • in-order traversal with callback and optional height and black height tracking for each node inspected (which allows for fast verification),
  • in-order ranged traversal, same as above,
  • breadth-first traversal with the same (simple dynamic FIFO queue implemented for this, fq.h/fq.c - two versions, pointer queue and data queue),
  • traversal callbacks
  • producing an ASCII dump of the tree (separate object to core rbt code)
  • optional pre-allocation of data of specified size (rbCreatePrealloc())

Example

Example usage (rbt_example.c):

#include <stdio.h>
#include "rbt.h"
#include "rbt_display.h"

static RbNode* callback(RB_CB_ARGS) {
	printf(" %u", node->key);
	return node;
}

int main(int argc, char **argv) {

	uint32_t count;
	RbTree *tree = rbCreate();

	for(int i = 0; i < 13; i++) {
		rbInsert(tree, i);
	}

	rbVerify(tree, RB_CHATTY, RB_FULL);
	char* d =
	rbDisplay(tree, 80, 11, RB_NO_NULL);
	printf("%s\n\n", d); free(d);

	printf("In order:");
	rbInOrder(tree, callback, NULL, RB_ASC);
	printf("\n");

	printf("Breadth first:");
	rbBreadthFirst(tree, callback, NULL, RB_ASC);
	printf("\n");

	printf("Between 4 (inclusive) and 9 (exclusive):");
	count = rbInOrderRange(tree, callback, NULL, RB_ASC,
		4, RB_INCL, 9, RB_EXCL);
	printf(", in range: %u nodes\n", count);

	printf("Between 4 (exclusive) and 9 (inclusive):");
	count = rbInOrderRange(tree, callback, NULL, RB_DESC,
		4, RB_EXCL, 9, RB_INCL);
	printf(", in range: %u nodes\n", count);

#if 0
	printf("Couldn't stop myself from adding this because of \n"
		"the increasing width of these lines of code here...");
#endif

	rbFree(tree);
	return 0;

}

Output:

$ ./rbt_example 
Valid red-black tree, node count 13, max height 5, black height 3

................................................................................
.......................................B3.......................................
................................................................................
...................B1......................................B7...................
................................................................................
.........B0..................B2..................R5..................R9.........
................................................................................
............................................B4........B6........B8........B11...
................................................................................
........................................................................R10.R12.
................................................................................

In order: 0 1 2 3 4 5 6 7 8 9 10 11 12
Breadth first: 3 1 7 0 2 5 9 4 6 8 11 10 12
Between 4 (inclusive) and 9 (exclusive): 4 5 6 7 8, in range: 5 nodes
Between 4 (exclusive) and 9 (inclusive): 9 8 7 6 5, in range: 5 nodes

Testing

Provided is a test program / benchmark, rbt_test.c. Options:

$ ./rbt_test -h
rbt_test (c) 2018: Wojciech Owczarek, simple red-black tree implementation

usage: rbt_test [-w NUMBER] [-H NUMBER] [-n NUMBER] [-r NUMBER] [-b NUMBER]
                [-s] [-m] [-e] [-l] [-o] [-i NUMBER]

-w NUMBER       Width of text block displaying the final tree, default 80
-H NUMBER       Height of text block displaying the final tree, default 20
-n NUMBER       Number of random nodes to insert into tree, default 1000
-r NUMBER       Number of nodes to leave on tree after removal, default 20
-b NUMBER       Number of nodes to randomly turn red to invalidate tree, default 0
-s              Test insertion only, generate CSV output on stdout
-m              Test removal only, CSV output to stdout
-e              Test search only, CSV output to stdout
-l              Test incremental search only (during insertion), CSV output to stdout
-o              Test decremental search only (during removal), CSV output to stdout
-i NUMBER       CSV log output interval, default every 1000 nodes,  unless
                1000 < 1% node count, then 1% node count is used.

Example output (mind that this ran on a shite Atom box, so performance is indicative of its shiteness):

$ ./rbt_test -n 1000000 -b 5
Generating 1000000 size random insertion, removal and search key arrays... done
Inserting 1000000 random keys... done
Verifying red-black tree... Valid red-black tree, node count 1000000, max height 24, black height 12
Finding all 1000000 keys in random order... 1000000 found
Finding all 1000000 keys in sequential order... 1000000 found
Performing in-order traversal with height and black height tracking... done
Performing in-order traversal without height and black height tracking... done
Performing breadth-first traversal with height and black height tracking... done
Performing breadth-first traversal without height and black height tracking... done
Destroying tree... done.
Re-adding 1000000 keys in random order... done
Removing all 1000000 keys in sequential order... done
Re-adding 1000000 keys in sequential order... done
Removing all 1000000 keys in sequential order again... done
Re-adding 1000000 keys in random order... done
Removing 999980 keys in random order to leave 20 keys... done

Test results:

+---------------------------------+-------------+---------+
| Test                            | result      | unit    |
+---------------------------------+-------------+---------+
| Insertion, count 1000000        | 1576        | ns/key  |
| Insertion, rate                 | 634395      | nodes/s |
| Verification, rate              | 6439571     | nodes/s |
| Search, count 1000000           | 1437        | ns/key  |
| Search, rate                    | 695772      | hit/s   |
| Seq search, count 1000000       | 242         | ns/key  |
| Seq search, rate                | 4118549     | hit/s   |
| In-order, with tracking, rate   | 6940080     | nodes/s |
| In-order, fast, rate            | 7751622     | nodes/s |
| Breadth first, tracking, rate   | 3464460     | nodes/s |
| Breadth first, fast, rate       | 7113365     | nodes/s |
| Destruction, rate               | 1906188     | nodes/s |
| Seq removal, count 1000000      | 272         | ns/key  |
| Seq removal, rate               | 3665748     | nodes/s |
| Seq insertion, count 1000000    | 536         | ns/key  |
| Seq insertion, rate             | 1865072     | nodes/s |
| Removal, count 999980           | 1903        | ns/key  |
| Removal, rate                   | 525460      | nodes/s |
+---------------------------------+-------------+---------+

Final tree with 20 nodes:

................................................................................
.......................................B12......................................
................................................................................
...................R5......................................B17..................
................................................................................
.........B1..................B7..................R15.................B19........
................................................................................
....B0........B3........B6........R9........B14.......B16.......R18.............
................................................................................
............R2..R4..............B8..B11...R13...................................
................................................................................
...................................R10..........................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................

Verifying red-black tree... Valid red-black tree, node count 20, max height 6, black height 3

Painting random 5 nodes red to invalidate tree... done

Most likely broken tree with 20 nodes:

................................................................................
.......................................R12......................................
................................................................................
...................R5......................................B17..................
................................................................................
.........B1..................B7..................R15.................R19........
................................................................................
....B0........B3........B6........R9........B14.......R16.......R18.............
................................................................................
............R2..R4..............B8..B11...R13...................................
................................................................................
...................................R10..........................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................

Red root violation
Red-red violation: key 5 red -> parent key 12 red
Black height violation: key 16 black height 1 != previous black height seen 2
Red-red violation: key 16 red -> parent key 15 red
Red-red violation: key 18 red -> parent key 19 red

Some benchmarks (worst-case / random performance)

Below are some plots taken from the CSV output for tests at different key insertion counts. This was done on a fairly decent Xeon box with 64G RAM. Duration measurement is done with a simple before/after clock_gettime(), which itself is non-instant (usually some 20 ns for a start/stop call pair with VDSO), so the more iterations per measurement, the closer the number is to "reality". There are some spikes which could be the CPU doing something else; I have not really investigated these. I could have passed these plots through a low-pass filter to produce nice, smooth log curves, but this shows the real performance (well, mostly - clock_gettime() can also produce spikes). Performance is clearly dominated by cache misses (and L2 / L3 cache size is also the source of the sawtooth-like patterns); that is not the point. What is important is that it is pretty clearly shown that the total time per insertion / deletion / search is a function of log2(n), and that search time is a significant contributor to both insertion and deletion. If the implementation was to be rewritten for top-down, the search and rebalance parts would have been combined, likely resulting in shaving off some / many cycles (TODO).

Method

Before a test runs, n-sized randomised insertion, search and deletion arrays of keys ranging from 0 to n-1 are generated using Fisher-Yates shuffle. Using premade shuffled sequences guarantees that an operation is successful every time, with no repetitions. Keys are inserted, searched and deleted by iterating over these arrays and duration of every k operations is measured and then divided by k. The incremental search during deletion (or decremental search) first finds k nodes from the randomised deletion array, then deletes them, and so on, in batches of k. The incremental search during insertion is non-optimal, because it generates a k-sized random sample of previously inserted keys every k, but it does so by fetching insert array indices from the search array, modulo dividing them by the current node count. There are definitely duplicates happening, but overall the behaviour is random enough - the lower-right corner plots are almost symmetrical. These incremental / decremental search times are subtracted from insertion and deletion times, and that is what constitutes the approximate rebalancing overhead - obviously any spikes are superimposed. Rebalance overhead function plots are pretty much flat (at higher node counts at least), which again is a known property of red-black trees.

Results

100k nodes:

rbt benchmark with 100k nodes

1M nodes:

rbt benchmark with 1M nodes

10M nodes:

rbt benchmark with 10M nodes

500M nodes*:

rbt benchmark with 500M nodes

1000M nodes (a billion):

rbt benchmark with 1000M nodes

Notes

I did not spend enough time analysing these plots, so for now the noise towards the end of the 1000M insertion cycle remains to be explained. Frankly I could not be arsed to wait for another 1e9 run, so as Officer Barbrady would say "IT COULD BE ANYTHING!!!".

All plots were made using kst2 / kst-plot, which remains my all-time favourite graphing and data analysis package, and likely a mark of my age and unchanging habits. While I have your attention - I have trawled through many papers, plots and benchmarks in my life, and let me say this: those who fail to specify axis descriptions and units, should not bother to publish plots at all and go back to school instead. K, thx, bye.

* - the OS clock in the server was left uncorrected during the 500M run and it was drifting badly enough to affect the perceived duration of each 20000-node batch; this is why the sustained search time is apparently decreasing for that run. It is not. This was corrected for the 1000M run. Why weren't you using CLOCK_MONOTONIC you say? I was. Duration measurement could probably be rewritten with rdtsc (TODO).

FAQ

Q: If this is for a dictionary index, why are you going with red-black trees and not hash tables? Surely you do not need in-order traversal or ranged searches for a dictionary?

A: Of course. Actually it is even worse, because this index will hold hashes of full tree node paths of a hierarchical configuration container. Implementing a non-shite dynamically resizing hash table however, is somewhat less trivial a task than implementing a red-black tree, and there is no amortisation with a red-black tree, there is no load factor and no emptiness either, even if temporary. I wrote this for convenience, not for ultimate performance, even though some gains could be made from top-down operations, and a k-ary tree could make great use of vectorisation. Plus, it this was a fun exercise. The aforementioned non-shite dynamically resizing hash table is next on my list, but let me tell you a secret: I am not a computer scientist. I am not a real programmer either!

Q: What is a red-black tree?

A: The "Black Diamond" variety of crape myrtle trees has a nice red-black appearance. Some decorative cherry trees are also good for that.

Q: No, a real red-black tree

A:

Drawing depicting a cartoon tree with red and black spots in places where limbs and branches divide, forming an n=32 red-black tree. A smiling sun is depicted above and a bird flying over the tree, defecating on it mid-flight.