Somebody said this to me recently...
These are ideas gathered from people around the world through books, in-person conversations and even self-reflection. I will be giving attribution to all the sources to the best of my ability.
Data warehouse vs Operational system
The Data Warehouse Toolkit : The Complete Guide to Dimensional Modeling by Ralph Kimball, Margy Ross
Data warehouse has a very different user requirement when compared to that of an operational system like the database system that records all sold items in realtime.
The main reason for this is that they are having two very different purpose. The data warehouse is existing as a business decision support system where as the operational systems are systems of record. This oftens leads to some very different technical requirements. And as a result, ideally operational systems are designed for fast writes and data warehouses are designed for fast reads.
Quick fix for a major memory leak in NiceGUI Python Library [as of March 2026]
Micael Jarniac on Github discussions
Memory leaks are often characterised by evergrowing demand for process memory inspite of the fact that in reality it is expected to be way lower. This is often due to issue with deallocation of memory that were used by objects, which were created in the usual flow of the program but for some flaw in the process design.
ui.run(reload=False, show=False)What is a Datawarehouse ?
Basic elements of the data warehouse from The Data Warehouse Toolkit : The Complete Guide to Dimensional Modeling by Ralph Kimball, Margy Ross
Consists of the processes and the storage within the entire system that takes data from the operational system all the way to the end business users. Therefore it comprises of the systems that extracts data from source systems, transforms it and , ideally loads that into a number of business process specific denormalized data stores called data marts and gives access into those high-quality data via various tools such as analytic applications, forecasting models, data mining applications, or using adhoc queries.
The sole purpose of this entire system is to support data backed decision making via querying of business data.
What a Datawarehouse ought to be?
Dimensional modelling Myths from The Data Warehouse Toolkit : The Complete Guide to Dimensional Modeling by Ralph Kimball, Margy Ross
It should not be just storing summarised data, in fact the most robust datawarehouses have the most granular data available in their data marts, which are the data stores that the end user querying tools have access.
Success should be based soley on the degree of user acceptance of the datawarehouse into the business decision making process. This may often take a querying experience for the end users which is quick and intuitive.
Data marts should be built such that they partition the entire data available to the end user access tools on the basis of business process, and certainly not on departments. This is because business processes - say orders - are often the true seams of the business data and will lead to one copy of data from the source systems and consistent dimesnion attribute values.