Depends on the type of coder, I've met too many old coders who try to keep the memory use low, performance high but code complexity is terrible because it's all one giant spaghetti ball of code.
For example now at work I've created a system which has a single master procedure( productionId, datasetId, stepId ) where NULL in the last two means all sets, all steps. I know some of the steps would be more efficient if merged, I know some contain one-time setup (but is hard to extract out) that's repeated many times when I run them on all datasets but for development it's a bliss. I can rerun a single step for a single set, a single step for all sets, all steps for a single set, I can easily time them (start and finish, per step, per set) and see what's making it choke not to mention if there's an error it's in a narrowly defined piece of code not the many-thousands-of-lines script it's replacing. A coworker of mine is starting to work on it setting up another production type and he loved the structure because it was so easy to grasp, even if he's only looked at a few steps.
Another feature I like is passing objects instead of values through layers. For example, say you have a form that has a string and a radiobutton but needs to have another UI element added, let's say a checkbox. If you pass the values as ( string, radioButton ) you have to change signatures everywhere. If you have an object FormValues, add the checkbox and pick up the value where it's needed. Is that efficient? Probably not, I guess I'm often passing ten values around when I only need two. But it saves a lot of pointless coding time when I find out that oh, I have to increase that from two to three. Defensive coding that makes it easy to expand or change functionality beats hardcoding every time.
I started out with a C64 which had 64kB of RAM, I'm not going to do that if we're talking about a million or a billion objects. But there are still people stuck in that mode where it's like every byte matters and it just doesn't. Make code that's easy to work with (verbose for clarity and descriptive names, but compact using standard functions and generic code where possible) and about 95% of the time it'll be worth more than trying to make it machine-efficient. A lot of "hardcore" developers dismiss abstractions as simplification for the simpletons and real developers code right on the metal, maybe not in assembler anymore but they kind of want to. It takes a real change of mindset to write code for coders, not code for the machine. Of course it must run in acceptable time with acceptable resource use, but that's often a low bar these days.