Hacker News Reader

Comment by jackmaney | Hacker News Reader

About three years ago, I was at a place that used Qlikview, and it was the worst software I've ever used in my life. It's technically able to handle larger datasets than Excel, but it's much slower than Excel at any kind of analysis or aggregation, it doesn't make use of SQL (even though the data it handles is tabular...instead, it uses a home-brewed, mangled, and inferior query language), and its visualizations and UI are sub-par compared to Tableau.

Good riddance.

View on HN

I think that's not terribly fair. I agree with you in part that it's got it's problems, as I had hours worth of data lost due to its stupid way of saving a single file completely only after a reload, and it's auto-backup barfed in some strange and unexplained way, but it's reasonably snappy at aggregating and allowing for drill downs.

I personally dislike the fact it relies on natural joins so you are forced to use defined columns with the same names and use Qlik syntax to disable table linking, but others love that feature.

As an ETL tool actually it's pretty good. But if they had decided to go with straight SQL and extended it to external sources then it wouldn't have stupid syntactical EXISTS and NOT EXISTS clauses, but that's all I can really fault it on.

It's a good tool for what it does, and it turns out that what it can do is a lot! The biggest issue I found was that whilst I picked up the syntax basics in a day and it's more intricate features in a week or two, most of the Qlik developers in Australia are morons who charge a lot and deliver very little. I was very lucky to have an amazing colleague who does know his stuff, but the consultants who we were eventually forced to get in were awful. Then again, the CEO of that company was just walked out the door for gross incompetence so it was probably a combo of stupid management and a limited talent pool for Qlik devs.

That's interesting, because I don't know what one should do to make QlikView slower than Excel. I used to do QlikView development full time for many years and I know pretty well what it's capable of, but I still can't imagine what that could be. QlikView is one of the fastest in-memory engines ever created. It's being perfected since 90s and now it's extremely optimized using techniques like JIT compilation, extensive parallelization with near linear scalability, and multiple layers of caching. Sub-second response times on datasets with tens of millions of rows on a rather average server are typical. It never came to my mind to compare it against Excel (since it's a totally different league), but I guess Qlik would be 3-4 orders of magnitude faster than Excel (PowerPivot is a different story). When you refer to QlikView's query language I guess you mean the Qlik load script which is used for initial loading data in memory. While it's indeed slow (it's single-threaded), it's never ever used for querying. It's not even technically feasible as user selections in UI can't trigger script.

QlikView does have disadvantages (mostly related to UI as you rightfully mentioned) but its performance is almost never among them.

One badly named column and you can easily setup a join that forms a Cartesian product - and you'd be none the wiser.

I hate their insistence on natural joins.

I can relate. The Qlik script is a half-baked product from 90s. After writing 10xKLOC in QLik script I decided to create a visual tool for data transformation (EasyMorph, http://easymorph.com) to not deal with any kind of scripting at all.

Assuming this sits in the same camp as Alteryx and the like?

Yes, and no. When compared to Alteryx we say EasyMorph does 50% of what Alteryx does for 10% of its price :) EasyMorph is more lightweight than Alteryx and is not positioned as a big data tool. Although in general it serves the same purpose with more focus on ad hoc data analysis and profiling.

Seems like a cool tool :-)

Your website gives a decent rundown, only part of it isn't displaying in portrait orientation on my iPhone6s+ :-(

I was in charge of supporting a large QlikView installation a couple of years ago. At one point we were just deploying servers with 512GB of RAM because the dashboards kept growing and growing and the developers were having trouble identifying the causes. Over time, updates helped but I was often in Process Monitor (SysInternals) reverse engineering the whole thing and suggesting fixes, grabbing memory dumps and analyzing stack traces, etc.

It was fun because it was a puzzle and I was helping fix broken things, but at some point you just ask yourself "what for?" when things are so broken and improve very slowly.

As the parent comment mentions, QlikView is now the previous generation product, and Qlik Sense is where they are moving towards.

Qlik Sense is pretty snappy, has much improved UX, etc.

Qlik are top-right quadrant for BI tools, on a par with Tableau.

Is it really though? I tried to find more info on Qliksense and perhaps I'm not looking hard enough, but I can't see a decent rundown of why it is superior.

Qlik Sense is about 10% faster than QlikView 11, although QlikView 12 got the same engine as Qlik Sense so they're on par now.

Reply on news.ycombinator.com