Yosh, I am back to blogging in the parallel world, with a whole new set of ventures to explore and bore you with the same. Okay, lets get down to the brass tacks straightaway.
I happened to stumble upon this exciting piece of software called TrailDB and allow me to explain to you my trails through it.
'Trails' in a database?
Before we find ourselves muddled in the technicality and perplexity of things, let us take a minute to note the primary objective of this innovative database - storing and querying time-based user events. If you are born and brought up with relational databases (like I have), you might be wondering, how is this any different? This should be your sweet and simple answer. Having highlighted that, I am not going further into the nuts and bolts of things; the documentation serves that purpose quite brilliantly.
The must-have installation quandary!
I hate installing stuff, especially when they say it is 'simple'. As the documentation promises, the installation on Linux is pretty straightforward. However, if you ever want to test your patience, I recommend installing the same on Windows (Use of bash is not an option). After fixing what seemed like a multitude of errors in the manual installation of dependencies, I realized with tears of grief, that the required version of libJudy (an interesting dependency) is unavailable for windows.
The biased documentation
When I finally decided to get my hands dirty with the actual API, it was all good till I got to the exciting part of running queries. I was using the python binding, but looking at C documentation! Nevertheless, it led me to several intriguing questions such as - what could be an equivalent of C struct in python? Dictionary? Tuple? List? Well, after staring at the source code for ample time, I deciphered the syntax to run queries ( 'event filters' in trailDB terminology).
The final verdict
At the end of the day, the assessment of any database comes down to two things - storage and query capabilities. While trailDB scores with the former (thanks to all the expensive stuff happening at 'C' level), the latter ain't all that impressive. For example, a simple query that involves comparison has to be done in the application world, as it can check only for equality! Not to mention, that the query has to be in conjunctive normal form only.
In my opinion, while it isn't suitable for applications that are database intensive, trailDB is apt if you are looking for a light database, say for archiving massive data (that can be represented as strings easily).
That brings us to the end of yet another exciting adventure in the parallel world. Relevant code will be available here. Adios and happy coding in a parallel world! 😊
I happened to stumble upon this exciting piece of software called TrailDB and allow me to explain to you my trails through it.
'Trails' in a database?
Before we find ourselves muddled in the technicality and perplexity of things, let us take a minute to note the primary objective of this innovative database - storing and querying time-based user events. If you are born and brought up with relational databases (like I have), you might be wondering, how is this any different? This should be your sweet and simple answer. Having highlighted that, I am not going further into the nuts and bolts of things; the documentation serves that purpose quite brilliantly.
The must-have installation quandary!
I hate installing stuff, especially when they say it is 'simple'. As the documentation promises, the installation on Linux is pretty straightforward. However, if you ever want to test your patience, I recommend installing the same on Windows (Use of bash is not an option). After fixing what seemed like a multitude of errors in the manual installation of dependencies, I realized with tears of grief, that the required version of libJudy (an interesting dependency) is unavailable for windows.
The biased documentation
When I finally decided to get my hands dirty with the actual API, it was all good till I got to the exciting part of running queries. I was using the python binding, but looking at C documentation! Nevertheless, it led me to several intriguing questions such as - what could be an equivalent of C struct in python? Dictionary? Tuple? List? Well, after staring at the source code for ample time, I deciphered the syntax to run queries ( 'event filters' in trailDB terminology).
The final verdict
At the end of the day, the assessment of any database comes down to two things - storage and query capabilities. While trailDB scores with the former (thanks to all the expensive stuff happening at 'C' level), the latter ain't all that impressive. For example, a simple query that involves comparison has to be done in the application world, as it can check only for equality! Not to mention, that the query has to be in conjunctive normal form only.
 |
| Some fancy compression! |
In my opinion, while it isn't suitable for applications that are database intensive, trailDB is apt if you are looking for a light database, say for archiving massive data (that can be represented as strings easily).
That brings us to the end of yet another exciting adventure in the parallel world. Relevant code will be available here. Adios and happy coding in a parallel world! 😊
Hey, I was wondering if I can use your image for a public presentation on TrailDB.
ReplyDeleteSure :)
Delete