We Keep Coding

SlidingWindows for slow data (big intervals) on Apache Beam

I am working with Chicago Traffic Tracker dataset, where new data is published every 15 minutes. When new data is available, it represents records off by 10-15 minutes from the "real time" (example, look for _last_updt).
For example, at 00:20, I get data timestamped 00:10; at 00:35, I get from 00:20; at 00:50, I get from 00:40. So the interval that I can get new data "fixed" (every 15 minutes), although the interval on timestamps change slightly.
I am trying to consume this data on Dataflow (Apache Beam) and for that I am playing with Sliding Windows. My idea is to collect and work on 4 consecutive datapoints (4 x 15min = 60min), and ideally update my calculation of sum/averages as soon as a new datapoint is available. For that, I've started with the code:
PCollection<TrafficData> trafficData = input
.apply("MapIntoSlidingWindows", Window.<TrafficData>into(
SlidingWindows.of(Duration.standardMinutes(60)) // (4x15)
.every(Duration.standardMinutes(15))) . // interval to get new data
.triggering(AfterWatermark
.pastEndOfWindow()
.withEarlyFirings(AfterProcessingTime.pastFirstElementInPane()))
.withAllowedLateness(Duration.ZERO)
.accumulatingFiredPanes());
Unfortunately, looks like when I receive a new datapoint from my input, I do not get a new (updated) result from the GroupByKey that I have after.
Is this something wrong with my SlidingWindows? Or am I missing something else?

One issue may be that the watermark is going past the end of the window, and dropping all later elements. You may try giving a few minutes after the watermark passes:
PCollection<TrafficData> trafficData = input
.apply("MapIntoSlidingWindows", Window.<TrafficData>into(
SlidingWindows.of(Duration.standardMinutes(60)) // (4x15)
.every(Duration.standardMinutes(15))) . // interval to get new data
.triggering(AfterWatermark
.pastEndOfWindow()
.withEarlyFirings(AfterProcessingTime.pastFirstElementInPane())
.withLateFirings(AfterProcessingTime.pastFirstElementInPane()))
.withAllowedLateness(Duration.standardMinutes(15))
.accumulatingFiredPanes());
Let me know if this helps at all.

So #Pablo (from my understanding) gave the correct answer. But I had some suggestions that would not fit in a comment.
I wanted to ask whether you need sliding windows? From what I can tell, fixed windows would do the job for you and be computationally simpler as well. Since you are using accumulating fired panes, you don't need to use a sliding window since your next DoFn function will already be doing an average from the accumulated panes.
As for the code, I made changes to the early and late firing logic. I also suggest increasing the windowing size. Since you know the data comes every 15 minutes, you should be closing the window after 15 minutes rather than on 15 minutes. But you also don't want to pick a window which will eventually collide with multiples of 15 (like 20) because at 60 minutes you'll have the same problem. So pick a number that is co-prime to 15, for example 19. Also allow for late entries.
PCollection<TrafficData> trafficData = input
.apply("MapIntoFixedWindows", Window.<TrafficData>into(
FixedWindows.of(Duration.standardMinutes(19))
.triggering(AfterWatermark.pastEndOfWindow()
// fire the moment you see an element
.withEarlyFirings(AfterPane.elementCountAtLeast(1))
//this line is optional since you already have a past end of window and a early firing. But just in case
.withLateFirings(AfterProcessingTime.pastFirstElementInPane()))
.withAllowedLateness(Duration.standardMinutes(60))
.accumulatingFiredPanes());
Let me know if that solves your issue!
EDIT
So, I could not understand how you computed the above example, so I am using a generic example. Below is a generic averaging function:
public class AverageFn extends CombineFn<Integer, AverageFn.Accum, Double> {
public static class Accum {
int sum = 0;
int count = 0;
}
#Override
public Accum createAccumulator() { return new Accum(); }
#Override
public Accum addInput(Accum accum, Integer input) {
accum.sum += input;
accum.count++;
return accum;
}
#Override
public Accum mergeAccumulators(Iterable<Accum> accums) {
Accum merged = createAccumulator();
for (Accum accum : accums) {
merged.sum += accum.sum;
merged.count += accum.count;
}
return merged;
}
#Override
public Double extractOutput(Accum accum) {
return ((double) accum.sum) / accum.count;
}
}
In order to run it you would add the line:
PCollection<Double> average = trafficData.apply(Combine.globally(new AverageFn()));
Since you are currently using accumulating firing triggers, this would be the simplest coding way to solve the solution.
HOWEVER, if you want to use a discarding fire pane window, you would need to use a PCollectionView to store the previous average and pass it as a side input to the next one in order to keep track of the values. This is a little more complex in coding but would definitely improve performance since constant work is done every window, unlike in accumulating firing.
Does this make enough sense for you to generate your own function for discarding fire pane window?

Java - Most efficient random-access multi-threaded list

Chosen List Structure:
Synchronised LinkedList.
Scenario:
My program requires rendering some (rather computational) generated images in a grid. These images must update whenever some data value changes (on another thread), hence, I have a rendering queue to manage this.
The rendering queue is a synchronised LinkedList, where on a low-priority thread, it is constantly being iterated over to check if some render work needs doing. Since the images are based on all kinds of data, each of which could change independently, I needed some form of queue to combine changes.
Data tends to change in chunks, and so when a large batch comes through I see an imaginary line run down the area where it's re-rendering the tiles. To pretty this up a bit, I decided rather than rendering in standard order, I'd render them in a random order (to give a 'dissolve in/out' effect).
It looks lovely, but the only problem is, there is a notable different in the amount of time it takes to complete with this effect running.
Problem:
I've theorised a couple of reasons accessing this list randomly instead of iteratively would cause such a notable delay. Firstly, the Random number generator's nextInt method might take up a significant enough amount of time. Secondly, since it's a LinkedList, getting the nth item might also be significant when the size of the list is in the 4000s range.
Is there any other reason for this delay that I might have overlooked? Rather than using a random number generator, or even a linked list, how else might I efficiently achieve a random access & remove from a list? If you've read the scenario, perhaps you can think of another way I could go about this entirely?
Requirements:
Multi-threaded addition to & modification of list.
Random access & removal of items from list.
Efficient operation, with large data sets & number of runs

You can use an ArrayList along with a couple of simple operations to implement this very efficiently.
To insert, always insert new work at the end of the list (an amortized constant time operation).
To extract a random piece of work, pick a random number i, swap the element at i with the element at the end of the list, and then extract and return that new last element.
Here's code (untested, uncompiled):
class RandomizedQueue<T> {
private final List<T> workItems = new ArrayList<>();
private final Random random;
RandomizedQueue(Random random) {
this.random = random;
}
public synchronized void insert(T item) {
workItems.add(item);
}
public synchronized T extract() {
if (workItems.isEmpty()) {
return null; // or throw an exception
}
int pos = random.nextInt(workItems.size());
int lastPos = workItems.size() - 1;
T item = workItems.get(pos);
workItems.set(pos, workItems.get(lastPos));
return workItems.remove(lastPos);
}
}

You could perhaps use a PriorityQueue, and when adding things to this queue give each item a random priority. The rendering can just always take the top element on the queue since it is randomized already. Inserting at a "random" position in a PriorityQueue (or better put, with a random priority) is really fast.

Existing solution to "smart" initial capacity for StringBuilder

I have a piece logging and tracing related code, which called often throughout the code, especially when tracing is switched on. StringBuilder is used to build a String. Strings have reasonable maximum length, I suppose in the order of hundreds of chars.
Question: Is there existing library to do something like this:
// in reality, StringBuilder is final,
// would have to create delegated version instead,
// which is quite a big class because of all the append() overloads
public class SmarterBuilder extends StringBuilder {
private final AtomicInteger capRef;
SmarterBuilder(AtomicInteger capRef) {
int len = capRef.get();
// optionally save memory with expense of worst-case resizes:
// len = len * 3 / 4;
super(len);
this.capRef = capRef;
}
public syncCap() {
// call when string is fully built
int cap;
do {
cap = capRef.get();
if (cap >= length()) break;
} while (!capRef.compareAndSet(cap, length());
}
}
To take advantage of this, my logging-related class would have a shared capRef variable with suitable scope.
(Bonus Question: I'm curious, is it possible to do syncCap() without looping?)
Motivation: I know default length of StringBuilder is always too little. I could (and currently do) throw in an ad-hoc intitial capacity value of 100, which results in resize in some number of cases, but not always. However, I do not like magic numbers in the source code, and this feature is a case of "optimize once, use in every project".

Make sure you do the performance measurements to make sure you really are getting some benefit for the extra work.
As an alternative to a StringBuilder-like class, consider a StringBuilderFactory. It could provide two static methods, one to get a StringBuilder, and the other to be called when you finish building a string. You could pass it a StringBuilder as argument, and it would record the length. The getStringBuilder method would use statistics recorded by the other method to choose the initial size.
There are two ways you could avoid looping in syncCap:
Synchronize.
Ignore failures.
The argument for ignoring failures in this situation is that you only need a random sampling of the actual lengths. If another thread is updating at the same time you are getting an up-to-date view of the string lengths anyway.

You could store the string length of each string in a statistic array. run your app, and at shutdown you take the 90% quartil of your string length (sort all str length values, and take the length value at array pos = sortedStrings.size() * 0,9
That way you created an intial string builder size where 90% of your strings will fit in.
Update
The value could be hard coded (like java does for value 10 in ArrayList), or read from a config file, or calclualted automatically in a test phase. But the quartile calculation is not for free, so best you run your project some time, measure the 90% quartil on the fly inside the SmartBuilder, output the 90% quartil from time to time, and later change the property file to use the value.
That way you would get optimal results for each project.
Or if you go one step further: Let your smart Builder update that value from time to time in the config file.
But this all is not worth the effort, you would do that only for data that have some millions entries, like digital road maps, etc.

How to generate incremental identifier in java

I have requirement in which I continuously receive messages that needs to be written in a file. Every time a new message is received it needs to be written in a separate file. What I want is to generate an unique identifier to be used as a file-name. I also want to preserve the order of the messages as well. By this I mean, the identifier generated as a file-name should always be incremental.
I was using UUID.randomUUID() to generate file-names but the problem with this approach is that UUID only assures randomness of the identifier but is not incremental. As a result I am losing the ordering of the file (I want file generated first should appear first in the list).
Approaches known
Can use System.currentTimeMillis() but I can receive multiple messages at same time stamp.
2.Another approach could be to implement static long value and increment it whenever a file is to be created and use the long value as a file-name. But I am not sure about this approach. Also it doesn't seem to be a proper solution to my problem. I think there could be far better solutions than this one.
If someone could suggest me a better solution to this problem, will be highly appreciated.

If you want your id value to uniformly rise even between server restarts, then you must either base it on the system time or have some elaborately robust logic that persists the last ID used. Note that achieving robustness on its own is not hard, but achieving it in a performant and scalable way is.
If you additionally need the id to be unique across multiple nodes in a redundant server cluster, then you need even more elaborate logic, which definitely involves a persistent store to which all the boxes synchronize access. Making this performant is, of course, even harder.
The best option I can see is to have a quite long ID so there's room for these parts:
System.currentTimeMillis for long-term uniqueness (across restarts);
System.nanotime for finer granularity;
a unique id of each server node (determined in a platform-specific way).
The method will still have to remember the last value generated and retry in case of a duplicate. It won't have to retry too many times, though, just until the next nanoTime clock tick—it could even busy-wait for it.
Sketch of code without point 3 (single-node implementation):
private static long lastNanos;
public static synchronized String uniqueId() {
for (;/*ever*/;) {
final long n = System.nanoTime();
if (n == lastNanos) continue;
lastNanos = n;
return "" + System.currentTimeMillis() + n;
}
}

Ok, my hands up. My last answer was fairly flaky and I've deleted it.
Keeping with the spirit of the site, I thought I'd try a different tac.
If you say you are keeping these messages in a single file then you could try something like creating an unique Id out of the size of the file?
Before you write the message to the file it's id could be the current size of the file.
You could add the filename + size as the id if these messages need to be unique across a number of files.
I'll leave the hot potato of synchronization to another day. But you could wrap all of this up in a syncronized object that keeps track of things.
Also, I am assuming that any messages written to the file will not be removed in the future.
ADDITIONAL NOTE:
You could create an message processing object that opens the file on construction (or via a create method).
This object will get the initial size of the file and this will be used as the unique id.
As each message is added (in a synchronized manner), the id is incremented by the size of the message.
This would address the performance issues. Will not work if more than one JVM/Node accesses the same file.
Skeletal Idea:
public class MessageSink {
private long id = 0;
public MessageSink(String filename) {
id = ... get file size ..
}
public synchronized addMessage(Message msg) {
msg.setId(id);
.. write to file + flush ..
.. or add to stack of messages that need to be written to file
.. at a later stage.
id = id + msg.getSize();
}
public void flushMessages() {
.. open file
.. for each message in stack write ...
.. flush and close file
}
}

I had the same requirement and found a suitable solution. Twitter Snowflake uses a simple algorithm to generate sortable 64bit (long) ids. Snowflake is written on Scala but the approach is simple and could be easily used in a Java code.
id is composed of:
timestamp - 41 bits (millisecond precision w/ a custom epoch gives us 69 years);
machine id - 10 bits (MAC address could be used as a hardware id);
sequence number - 12 bits - rolls over every 4096 per machine (with protection to avoid rollover in the same ms)
Formula looks like: ((timestamp - customEpoch) << timestampShift) | (machineId << machineIdShift) | sequenceNumber;
Shift for each component depends on it's bits position in ID.
Detailed description and source code could be found at github:
Twitter Snowflake
Basic Java implementation of the Snowflake algorithm

Transforming code to multithread version

I have the following problem:
for example, I have 10 lists, each one has a link to some other lists, I would like to create a code to search elements in theses lists, I've already done this algorithm but sequentially, it start the search in the first list, then if the search is failed, it send messages for searching in lists that have a link with it (to the first one), at the end of the algorithm, he show the results as the number of lists visited and if he find the element or no.
now, I want to transform it to be a parallel algorithm, at least a concurrent one using multi-threads:
to use threads for searching;
to start a search in the 10 lists at the same time;

As long as you don't change anything, you can consider your search read only. In that case, you probably don't need synchronization. If you want to have a fast search, don't use threads directly but use runnables and look for the appropriate classes. If you do work directly with threads, make sure that you don't exceed the number of processors.
Before going further, read into multi-threading. I would mention "Java Concurrency in Practice" as a (rather safe) recommendation. It's too easy to get wrong.

I am not sure from your problem statement if there is a sequential dependency between the searches in different lists, namely whether search results from the first list should take priority over those from the second list or not.
Assume there's no such dependency, so that any search result from any list is fine, then this is expressed in a very concise way as a speculative search in Ateji PX:
Object parallelSearch(Collection<List> lists)
{
// start of parallel block
[
// create one parallel branch for each of the lists
|| (List list: lists) {
// start searching lists in parallel
Object result = search(list);
// the first branch that finds a result returns it,
// thereby stopping all remaining parallel branches
if(result != null) return result;
}
]
}
The term "speculative" means that you run a number of searches in parallel although you know that you will use the result of only one of them. Then as soon as you have found a result in one of the searches, you want to stop all the remaining searches.
If you run this code on a 4-core machine, the Ateji PX runtime will schedule 4 parallel branches at a time in order to make the best use of the available parallel hardware.