Prolog Business Cases



To establish the use of Prolog in an organization, management must be made aware of the immense benefits it has from a business perspective, such as: In any organization, there are countless possible applications for Prolog. In the following, we consider some of them.

Prolog as a database

One of the most basic use cases of Prolog is to rely on its ability to act as a database.

By this, we mean applications that simply store data and make it available for querying.

Prolog is very well suited for such applications, since you can dynamically add and remove data, and rely on automatic indexing to obtain good performance in many situations of practical relevance. Also, you can easily edit and process Prolog programs by various means, which makes exporting, importing and processing the accumulated knowledge in general very convenient. In addition, formulating Prolog queries is often more natural and easier than formulating SQL queries. For such reasons, Prolog is often a good alternative for other database systems.

There is a syntactic subset of Prolog that is called Datalog, and it is especially suitable for such use cases.

Analyzing events and anomalies

A logfile records events. For example, the Prolog HTTPS sever that sent you this page keeps a logfile that describes each request it processes.

When I locally test the server by fetching the page /prolog, it writes an entry similar to the following in its logfile:
/*Sun Apr  9 23:01:19 2017*/
request(45, 1491771679.213,
         [peer(ip(127,0,0,1)),
          method(get),request_uri('/prolog'),path('/prolog'),
          http_version(1-1),host(localhost),port(3037),
          connection('keep-alive'),upgrade_insecure_requests('1'),
          if_modified_since('Wed, 05 Apr 2017 23:24:08 GMT')]).
    
This entry describes the request of the client.

After processing the request, the server automatically writes an additional entry that describes how the request was processed. For example, in my case, the following entry was added:
completed(45, 0.0009790000000000076, 0, 304, not_modified).
    
The key point is that this logfile is a series of Prolog facts. Hence, the whole logfile can be easily analyzed using Prolog!

A single invocation of Prolog suffices to load the entire file, and at the same time makes it ready for posting arbitrary queries over the accumulated data.

For example, in the case above, we can ask: Are there any client requests that were not served? In Prolog, this becomes:
?- request(R, _, _), \+ completed(R, _, _, _, _).
false.
    
In our case, this means that the server at least handled each request in some way, and did not silently ignore one.

Of course, not all logfiles can be parsed directly with Prolog. However, the more you use Prolog syntax within your organization, the more you will benefit from being able to post queries over such data. In addition, you can typically easily convert any logfile format to valid Prolog facts, and then apply the same reasoning.

Rule-based reporting

In every organization, there is a constant need for excellent reports. Prolog is ideally suited for generating such reports from available data.

For example, imagine you are responsible for reporting the number of monthly visitors of your company's homepage. Further, for historic reasons, your team of technicians sends you the average number of visitors per day, at the end of each month. That is, the data you have available is stored in the following convenient form, using Prolog facts, indicating a nice growth of daily visitors to your site:
year_month_daily_visitors(2017, jan, 30018).
year_month_daily_visitors(2017, feb, 32913).
year_month_daily_visitors(2017, mar, 35871).
    
However, your management wants you to report the total number of visitors for each month, not the daily average.

Using Prolog rules, it is straight-forward to convert your data to the data that your management wants to see. For example, using integer arithmetic, we can readily relate the daily average to the monthly total:
year_month_total(Year, Month, Total) :-
        Total #= Days*Daily,
        year_month_days(Year, Month, Days),
        year_month_daily_visitors(Year, Month, Daily).
    
This only requires the definition of year_month_days/3, with clauses that may look as follows:
year_month_days(2017, jan, 31).
year_month_days(2017, feb, 28).
year_month_days(2017, mar, 31).
    
This suffices to produce the answers we need, for example:
?- year_month_total(Year, Month, Total).
Year = 2017,
Month = jan,
Total = 930558 ;
Year = 2017,
Month = feb,
Total = 921564 ;
Year = 2017,
Month = mar,
Total = 1112001.
    
Using format/2, you can turn this into a more readable report:
?- format("year month visitors~n", []),
   year_month_total(Year, Month, Total),
   format("~w ~w ~t~D~20|~n", [Year,Month,Total]),
   false.
year month visitors
2017 jan     930,558
2017 feb     921,564
2017 mar   1,112,001
    
Notice the use of forced backtracking via false/0.

A few weeks later, management needs a different report: They now want neither the daily average, nor the total number, but rather the cumulative sum of visitors for the current year, up to and including the current month.

Using Prolog, it is straight-forward to generate this report by simply formulating the required rules, while retaining the original data exactly as it was. For example:
cumulative(Pred, Until, C) :-
        months(Ms),
        cumulative_(Ms, Pred, Until, 0, C).

cumulative_([], _, _, C, C).
cumulative_([M|Ms], Pred, Until, C0, C) :-
        C1 #= C0 + Current,
        call(Pred, M, Current),
        if_(M=Until, C=C1,
            cumulative_(Ms, Pred, Until, C1, C)).

months([jan,feb,mar,apr,may,jun,jul,aug,sept,oct,nov,dec]).
    
This uses the meta-predicates call/3 and if_/3 to describe the cumulative sum in a very general and flexible way. For example:
?- cumulative(year_month_total(2017), M, C).
M = jan,
C = 930558 ;
M = feb,
C = 1852122 ;
M = mar,
C = 2964123 ;
false.
    
As a special case of this query, we can of course also inquire any particular month:
?- cumulative(year_month_total(2017), feb, C).
C = 1852122 ;
false.
    
Note the massive advantages of such a way to reason about your data. For example, it suffices to send around Prolog facts, which you can easily do via plain text. Augmenting these facts is also very easy: It suffices to add one fact per month to keep the above report current.

Strategic considerations

Suppose you are successfully applying Prolog within your company to solve difficult problems. Would you publicly announce the fact that you are using Prolog, so that your competitors can benefit from the same technology? My personal guess is: No, you wouldn't.

Sometimes, when you introduce Prolog in an organization, people will dismiss the language because they have never heard of anyone who uses it. Yet, a third of all airline tickets is handled by systems that run SICStus Prolog. NASA uses SICStus Prolog for a voice-controlled system onboard the International Space Station. Windows NT used an embedded Prolog interpreter for network configuration. New Zealand's dominant stock broking system is written in Prolog and CHR.

A commercial Prolog system easily costs thousands of dollars. This is not software you "just buy" as an individual. A professional Prolog system is typically bought by companies who simply need the power of Prolog to solve their tasks. They are unlikely to advertise their internal technologies. Still, a few of them do. For example, the Java Virtual Machine Specification contains a lot of Prolog code, and IBM Watson uses Prolog for natural language processing.

Sometimes, instead of trying to introduce Prolog in an organization, it is more efficient to start working for one where it is already being used!


More about Prolog


Main page