As many of you know, I have been at this for a while. So from my back pages comes this.
Early on, we watched every byte of the program executable and managed the executables at the register level. If one section of the program was too long, we would allocate a new pseudo-register and code in the new section. In addition, we really did manage down to how a binary bit field would be declared and how to and or or against the bits.
Later when COBOL and that generation of languages came to be, we argued how to use the search and searchall subroutines, self-defining datatypes, and using the built-in sorting or writing your own bubble sort. There were intense arguments as to what was faster and cleaner.
In the next generation of code, we were told not to worry about the efficiency of the code. The folks who paid us understood that developer costs were the most costly so who cared how you wrote it just get it done and if it was slow hardware was cheap compared to developer time so they just bought more memory, disk, etc. That lasted a while and then it started to catch up to them and the cost for hardware was very high.
Soon the efficiency in the backend became an issue and the RDBMs became king. With indices, being managed internal to efficient firmware and the ability to multiprocess on the backend the industry was able to regain some of the lost efficiency.
However, the need for IT processing has exponentially increased over time, and it continues to grow more rapidly them we imagine. With data growing exponentially and the need or desire for that data increasing almost as fast as the mass of data increases we find that efficiencies need to be captured anywhere we can find them.
I really do like the idea of reuse as mentioned earlier. However, elegant code and reuse is only useful to the developer. There are times is will be sacrificed for the needs of throughput the user demands. This is the tradeoff. We can be elegant if it is efficient. Elegance matters some time only to the developer. Speed and efficiency matters at all times to the user.
Therefore, the pendulum has swung again to the efficiencies, and we are again to understand that this is an art, and a science but we need to be quick about it.
Not all gray hairs are Dinosaurs!