Hacker News Reader

Comment by ninkendo | Hacker News Reader

What’s a SAS

Just five years back, this question would've been sacrilege in the 'data analytics' community.

5 years ago, I worked at a company that did a lot of data analytics stuff (though I did not; I was responsible with some web applications). They swore by SAS. To read the comments here are pretty jarring, considering that previous history and how entrenched they seemingly were. I remember conversations with the SAS users and saying if I were to go in that arena, I'd most likely be a bigger fan of using R which seemed like sacrilege in the context of the conversation.

Granted, I worked about 30 minutes from SAS's HQ in Cary, but still.

Kind of like COBOL, an old-school legacy thing originating from mainframes that some people still use.

Most of the worlds largest enterprises run tons of this legacy cobol mainframe stuff. They solve really hard problems using this technology and we would be wise not to be dismissive but instead to learn.

Right they do, but they are also all working really hard on their transition plans away from it. See the original link for concrete data on that.

It depends. I work professionally with Fortune 500 companies and for many of them mainframes are here to stay. It’s true that many of them wish to modernize and move away from mainframes; but they don’t want to get rid of them entirely.

Do people solve really hard problems with COBAL? I thought it was primarily a business language that replaces a lot of manual work. What sort of things has it found a niche for? Similar to how FORTRAN is used a lot in big data/physics?

Every bank, insurance, airline, and electric utility in the world older than 25 years old has a mountain of critical, working, and legacy software running COBOL and FORTRAN.

With the generation of programmers who wrote the code retiring it’s even more entrenched because even though they rely on that code dearly, few if any people remember why it does what it does how it does it.

> With the generation of programmers who wrote the code retiring

For many big COBOL systems that’s long past; in many cases the programmers that were coming on board when that generation retired are near retirement.

I've often thought it's probably a cracking time to be get some COBOL skills - you could jack your day rate up to the point you could probably name your price. There's an institutional incentive to replace it eventually, so not a long term thing, but you could probably milk it for a good few years.

I know COBOL programmers whose day rate is considerably north of £1000

That doesn't address the question I was asking, of course there are a lot of legacy COBOL systems out there.

Do any of those legacy systems work on "hard problems" similar to how FORTRAN is still used in fluid dynamics, physics, crystallography and chemistry?

Where would one look if one wanted to learn COBOL though? There's a free C implementation, but barely anything else surrounding it.

Where would one look if one wanted to learn COBOL though?

One would look at google. https://www.tutorialspoint.com/cobol/index.htm

We use SAS Enterprise Guide as a Business Intelligence/analytics tool at my work (Industrial Plant) - most of our production reporting is SAS based. We also use it for adhoc analysis/troubleshooting. I.e to spot long term declines in yield and that sort of thing.

Where SAS succeeds is they make it very easy for non programmers to do things like query databases extract, transform, merge data, create computed columns and output graphs and reports.

Enterprise Guide is 100% scriptable and has a programming language underneath it the "SAS langauge" basically anything you can do in the GUI has code that is generated in the backend which user can edit directly if they desire. In old versions of Enterprise Guide code the GUI generated tended to be pretty inefficient (especially the SQL side of things) so a lot of people hacked the code directly and bypassed the GUI - nowadays it has gotten pretty good not as much need to hand optimize things.

Best analogy I can give is Enterprise Guide is like Excel on steroids.

SAS also has very powerful statistical capabilities beyond what something like Excel offers. We leverage a lot of this stuff when we need to build engineering models, things like Time Series Analysis, Statistical Process Control, Linear and Non Linear Modelling (Including Stepwise and Logistic Regressions). There is even some Optimization solver type stuff I've used in past (though MATlab is more often used in my org for this work). SAS kind of takes the kitchen sink approach in R you need to install external packages from CRAN to unlock all this stuff in SAS tends to be "baked in". I've found if I read about a statistical technique in a paper good chance it is available in SAS - good example is a few years ago a grad student came to our plant and helped us perform some analysis using a technique called "Projection to Latent Structures" (PLS) to identify relationships between various factors on plant causing yield loss. At the time this wasn't very common - I think it is used in machine learning nowadays but SAS had a PLS procedure available for us to use (kind of like a function in other programming languages) and the documentation SAS provided was very good (especially for someone like me with engineering background rater than stats background).

The only other tool I know of that is remotely similar to Enterprise Guide is IBM's Cognos software and that is vastly more limited as far as the statistical modelling features are concerned.

My work is keen to try other software, in part due to license costs, I think. In the Last 6 months I've been on two training course for Microsoft Power BI, and Microsoft R-Open (which I think is just Open source R-Studio with some extra proprietary stuff thrown in).

I've now used both tools to do my day to day job. I managed to get by with each of them but I found usability fell short of ease of use of enterprise guide.

Power BI was nice and I think it has something to offer as far as data visualization but you need to do the actual modelling in some other language so doesn't really eliminate needs for SAS - more like compliments it.

R on the other hand is a lot lower level it has all the advanced statistics but you have to write the low level code yourself to unlock them which requires learning an entire programming language to build simple models. It will be a pretty hard sell getting Engineers (I'm talking mechanical, chemical engineers not software kind) and finance analyst people to use a tool like R.

I think R would hugely benefit from something like an "Enterprise Guide for R" - I.e a drag and drop GUI which generated R code for you.

Also I wouldn't say the SAS Languge itself is strictly an old-school legacy language (sure the origins of SAS are from the 70's and 80's and there is a lot of legacy cruft in the SAS programming language I'll give you that) - For me SAS is a little like C++ there is a heap of stuff that lives in the langauge for backwards compatibility with old programs but modern C++ program looks very different from early 90's C++ - same with SAS code.

I've found modern ODS graphics (SGplot, SGpannel, etc) was very much on par with "ggplot2" functionality in R which is what the training course covered for producing reports using R.

SAS to me is a bit like MATLAB (and hey we use both in Engineering world). Everyone hates this software and wants something better but nothing else approaches the usability of these tools so models continue to be built using them.

Not sure if you've heard about Alteryx, which imo comes a lot closer to SAS. Alteryx is like SPSS for 2017. It's fast, easy to use, drag and drop gui and works well in Production.

Alteryx looks cool. Except for the pricing, which is mind-blowingly ridiculous.

Thanks for this very informative interesting write-up, notes such as these are what makes Hacker News so great. I’m curious, what do you think of SQL? I come from a finance background was an Excel guru but now prefer PostgreSQl, amazing what you can do with their Windows Functions, Recursive CTEs, Lateral Join Set Functions, Custom Aggregates, also has Python and R built in. I don’t like PowerBi too disjointed from Excel and v weak SQL support why does Microsft have to keep imposing their own languages on us instead of alloing us to use far superior SQL, would be awesome to have SQL in Excel (python Exel wings offers this but limited and clunky)Main proble with PowerBi for me is you can’t update the database without obtuse hacks.

You can run R in SQL Server and Excel has the Data Model/PowerPivot now.

I used to work in a large international bank where SAS was the standard and while it can be used you can really feel the age of the technology.

Now I work in a startup bank and since I could chose the tools we're now running PostgreSQL/R/Python (RStudio and Spyder). Lately the marketing department are also using Tableu.

It's another world when it comes to output and efficiency of the team I lead compared to using enterprise guide.

Where I currently work, we teach all of our BAs (regardless of how technical they start out), to use SQL (with redash), and some of them use Tableu. As this approach has been running quite smoothly for us, would SAS offer anything that a team like ours might be missing?

Yeah I thought - Special Air Service or SaaS

Or the Scandinavian Airline

Surfers Against Sewage!

In my defence, I'll simply note that there's history here:

https://groups.google.com/forum/#!topic/comp.soft-sys.sas/oF...

https://www.youtube.com/watch?v=2v1m-75I3_o

Both actors appearing in that clip are total muppets but it seems relevant ;-)

Serial Attached SCSI?

Reply on news.ycombinator.com