#91: ABAP+VSC, SAP Cooking, RPT-1

Special Breaking News Announcement

We’re excited to welcome the famous author and ABAP expert Paul Hardy as a contributor to this issue! Paul is renowned for bestselling ABAP To The Future books and, most importantly, for the sensational SAP Opens Brewery blog post. Smash that Like button extra hard for Paul!

In this issue:

• Cooking With SAP by Jelena

• Beware the IDEs of March by Paul H.

• A Rapid Report on RPT by Paul M.

• Getting Data From SAP by Jelena

• Performance Reviews: Just Why? by Jelena

• Streamlit by Paul M.

Cooking With SAP

(Just one of my cookbook shelves IRL, awaiting sponsor endorsements)

If SAP was Food Network, then TechEd would be Iron Chef: highly anticipated, charged with expertise, amusing. But ultimately, after watching Masaharu Morimoto or Cat Cora cook the heck out of that squid or pheasant, we would go back to a spaghetti-and-meatballs dinner.

After SAP has unveiled new ingredients and gadgets with such panache, we are expected to cook with them. In cooking, we have centuries of experience and solid knowledge of what works and what doesn’t (“all mushrooms are edible but some only once”). We have trained at the elbows of our grandmas, babushkas, abuelitas, and Omas. In technology, we don’t have the same luxury; it’s always changing, always new (or at least appears that way).

In the SAP world, we need our food bloggers like Smitten Kitchen or Sally’s Baking (these ones not as much though). And there can never be too many cookbooks, that’s just not possible. Yet the SAP equivalent of Salt, Fat, Acid, Heat remains unwritten for some reason.

The sources we do have remind me of the main Soviet cookbook: The Book of Delicious and Healthy Food. It was notorious for its ridiculously vague instructions, such as “boil until done” (without any scoop on what “done” means), and recipes with some boujee ingredients you’d never see at a store in your life. That book traumatized generations of Soviet children.

Experimentation is also big in cooking. I’m glad that this concept finally reached the Iron Chefs of SAP and the unveiling of RPT-1 came with a Playground. Maybe we could reopen the issue of the real BTP free tier too?

Cooking with SAP is not always easy. But like some foodie addicts, we just can’t help it. Must. Keep. Cooking. So start your engines, chefs, and let’s find which cuisine reigns supreme! JP

Beware the IDEs of March

At SAP TechEd 2025 it was announced that SAP had done a bit of a 180-degree pivot, that is instead of recommending ADT (ABAP in Eclipse) now it is going to be ABAP in VS Code, supported by SAP.

There is nothing new under the sun. Many moons ago the plus for ADT was that you could do ABAP and UI5 in the same IDE. Then you could not. The UI5 stuff had to be done in the BAS which was an “improved” version of VS Code.

Now you can do both ABAP and UI5 again in the same IDE. Hooray!

The justification for the change in position is that AI works so much better in VS Code than Eclipse. I bet the Eclipse people have something to say about that but let it go.

I have always liked the look and feel of VS Code better than Eclipse, but that is just a beauty in the eye of the beholder type of thing.

What I will say is that for many years now it was possible to do your ABAP development in VS Code due to assorted community plug-ins. The connection to your back-end SAP system worked 100% the same as ADT and you even got the “quick fixes”. Moreover, you could also install the exact same plug-ins that BAS uses for UI5 development.

I will end with two million-dollar questions, neither of which I have the answer to as yet.

1. Will the official SAP version of ABAP in VS Code have all the features of the community-based version?

2. Will SAP charge ten billion dollars for this, as opposed to the community version being free?

Cheersy Cheers. PH

A Rapid Report on RPT

At TechEd events this year, SAP made some noise about a new kind of AI model they’d created: SAP-RPT-1. RPT-1 (I think they’re going to try to get people to say “rapid”) is based on the same neural network architecture as all the big famous AIs, the Transformer. But it’s not built for text: RPT-1 helps you “gain predictive insights from enterprise data without building models from scratch”. It wants to be a shortcut for all those prediction models that data nerds were squawking about before chatbots became the headline thing in AI.

As the Boring Enterprise Nerds’ Kool-Aid drinkingest AI dork, I want to give you a few other trivia tidbits you might find interesting about RPT-1.

• The tokens, unlike LLM tokens, are not pre-set words and word-chunks. With ChatGPT, you can know ahead of time all the “words” it knows by listing out the values in its tokenizer. RPT-1 isn’t like that. Its tokens are database cells, a combination of table name, column, and the value inside the cell. So when you use RPT-1 to predict something, you can know exactly how many tokens it will cost just by counting up rows and columns.

• At 22 million parameters, it is at least a thousand times smaller than any of the well-known large language models. 10-50 thousand times smaller than the big ones. So pay attention to its pricing when it goes into paid mode, because it should be FAR less expensive to run.

• The paper claims accuracy at 93% of a specially-trained purpose-built model for classification tasks, meaning that with one model SAP is saying you can get something that works 93% as well as a giant pile of specialist models.

SAP is starting out with RPT-1 by offering it in Generative AI Hub, but you can try it yourself at the playground. I haven’t heard any news about RPT-1 directly moving into SAP products. That would be a signal that SAP really believes in its power. PM

Getting Data From SAP

To borrow from Goodfellas: as far back as I can remember, people always wanted to get data out of SAP ERP and use it someplace else. For outsiders looking in, our SAP garden must seem like a bizarro world. Like this Redditor, they wonder: “If I use Power BI to connect to a SQL database and get the data for my dashboard, why can’t I just do that with SAP?” Oh sweet summer child…

Today, there are certainly more ways to get to your corporate SAP ERP data out than there were two decades ago. And even in the latest SAP TechEd keynote, “making SAP more open” and “every enterprise is becoming a data company” (whatever the heck that means) got a mention.

SAP Mentor and BI expert Ronald Konijnenburg recently shared two excellent LI posts on the subject. His article The Great Data Liberation focuses more on the technology side and offers a great “state of SAP data realm” summary. But his more obscure post talks about the thornier issues of compliance and licensing. The latter is frequently unknown to the “connect to SQL DB” crowd and definitely doesn’t make getting data from SAP any simpler. JP

Performance Reviews: Just Why?

Folks, it’s that time of the year again. You know it, you dread it. The annual performance review. Ugh.

So, let me get this straight. Instead of discussing something with the manager as and when it happens, we are to wait till end of year to debrief? “Oh, you’ve done messed up, Timmy. Let’s talk about it in December!”

And the requisite punch in the gut aka “opportunity for improvement”. What could I have done better? Nothing! If I could’ve done better, I would have.

Google “performance review” and the top results would be about how bad and dumb that practice is. My favorite quote: “if performance reviews were a drug, they would not meet FDA approval for efficacy”. Indeed. What is the typical performance review result? “Here is your annual inflation adjustment, there isn’t more money in the budget regardless of how you worked all year.” Talk about adding injury to insult. JP

Streamlit

This story is pretty simple: you should really try Streamlit. Streamlit is a quick and powerful Python framework to create data-friendly web apps. The linked video above is most of the tutorial you’ll really need, because any AI assistant code helper already knows Streamlit very well. You can start with a .csv of some data you want to wrangle and show off, and in minutes you’ve got a web thing that’s easy to share.

I think the only trap is the ease of getting from nothing to something. You should view Streamlit as prototype and show-off friendly, but not as big-honking-app friendly. A couple years ago I made the mistake of showing some cool ideas to folks, and by the time I thought “you know, this would scale a lot better with something else…” it was too late. I was stuck with something that people loved but was groaning under the weight of ambitions too large for the framework to carry. Stick to visualizing data and simple use cases, and Streamlit will take you far. PM

• The Nerdcast is back with exciting episode SAP TechEd Reality Check. The esteemed guests, book writers Paul Hardy (also a guest author in this issue!) and James McDonough stopped by. Watch, like, subscribe!

• Let the record state that we were ahead of time with the ABAP and VSC Nerdcast episode featuring Lars Hvam Petersen. How have the turntables turned!

• Are you interested in sponsoring the Nerdletter and featuring your amazing product? Have a story to share or feedback to give? Get in touch!

• This newsletter is always free and proudly written by humans and for humans. Show your support to resist AI slop!