Display of Stack-Trace Statistics in daily runs and production environments: 

Introduction: Log Analysis for error detection and reporting is an important activity for any commercial deployment. Together with powerful log scan and search techniques, log analysis both manual and automated are useful for the smooth running of any environment. Many in house and out of box commercial log analysis tools come with rich interfaces in the form of query language operators as well as dashboards to help the end users. However, the use of these query tools and their frequency is left to the discretion of the user. Most investigations based on log are on a case by case basis and therefore do not cover the overall hashes and counts of exception stack-trace. This write-up introduces automated stack-trace hasher from log scans as a useful addition to any statistics. 

Description: The conventional mode of query displaying graphs based on exception introduces the notion of filters that select exceptions matching one or more attributes usually in the form of keywords. Existing dashboards then serve up these selections in the same manner that they do with all exceptions on a continuous timeline.  

On the other hand, exceptions occurring in one code path often manifest themselves more than once. Depending on the frequency, these exceptions and their count indicate health of the component in the service that is logging these exceptions. While the dashboard for archived exceptions maybe available to view with pan and zoom, generally the statistics between service upgrades are missing. These can be easily solved with exception stack-trace hashing and taking snapshots of their running counts between upgrades and including it in the summaries. Such health indicators will be very helpful even in test runs of the service in various environments. Moreover, unnoticed exceptions can now get visibility and defect tracking may open up defects against the component based on these exceptions  

Design:  

 As the exceptions arrive, they are hashed and counted. This results in a frequency table that then gets displayed. We utilize both the exception frequency and inverse exception occurrence lookup for best usage.  

Languages: Sample stack trace hashing utilities are also available in Python and Java such as exhash. Therefore, taking a set of class, methods and optionally offsets as frame entries and creating a hash from a stack-trace using only class and methods will help with this tally. 

Conclusion: Utilizing a stack trace hasher in addition to test run reports or service monitoring will be helpful as a health indicator for the different components in a service. 

#codingexercise

Find the Newman-Shanks-Williams prime

This is defined by the recurrence relation:
F0 = 1
F1 = 1
Fn =  2 * Fn-1 + Fn-2
double GetNSWPrime(double n)
{
    if (n < 0) 
        return 0;

    if (n == 0 || n == 1)
        return 1;

    return 2 * GetNSWPrime(n - 1) + GetNSWPrime(n - 2);

}

>>> def p(n):
...    if (n==0) or (n==1):
...       return 1
...    return 2*p(n-1)+p(n-2)
...
>>> p(3)
7
>>> p(5)
41
>>> p(7)
239
>>>
Prime numbers initially are 1 or 2  units apart. The recurrence happens to be defined only on n-1 and n-2.