Hacker News Reader

Comment by megous | Hacker News Reader

In all seriousness, what's so noteworthy about this? It's one trivial 50 line recursive function. This is kind of stuff devs just type out instead of searching for a npm package, because it's faster that way.

View on HN

You only see the finished product, not the 100 failures and incremental improvements over time. You could certainly write something yourself from scratch, but I doubt it would be competitive against this.

Mmm, agree with parent; can get faster. Unfortunately lots of perf critical code is created out of necessity and defaults to proprietary.

You could profile this lib and figure out where it’s spending most of its time and figure out if there any ways to improve (starts getting into implications of JS <-> system), but as you get more idiosyncratic you lose flexibility and ease of use.

Eg type checks, heap additions, recursion, etc can get “slow”

> not the 100 failures and incremental improvements over time

This code has 3 actual code commits over 6 days, none of which changed anything fundamental. What are you talking about?

It's quite possible to compete against this on multiple levels. One would be correctness. The other a bit less wasteful interface. Another would be handling arrays in a more useful way (or at all). And yet another would be safety. Another would be speed and memory use. Yet another would be non-recursive implementation. Another would be documentation (what's this diffing? enumerable/non-enumerable properties? how it deals with getters? how about inherited properties? - I can't know any of this without looking long and hard at the code)... etc. etc.

Note on only the first part: the author might have tested bunch of stuff on their local without commiting until they've found something worth, perhaps? That would result in very few commits.

Nah, this just really isn’t a very complex or noteworthy library. There’s not that many edge cases to test. Just look at the test file.

>Microdiff aims for working for 99% of use cases, instead of making every edge case work.[1]

That's fine, of course, but probably should be in the README.

[1] https://github.com/AsyncBanana/microdiff/issues/2#issuecomme...

I wouldn't be able to write this. I mean, I could, but I lack the knowledge of Typescript and willingness to dedicate the needed time to create this while I have got other things to do.

The code is clean, well organized and nice to look at, easy to read through and understand, and it's great that someone has made the effort to create this and share it for us to use.

Would you rather implement `leftPad` or use the npm package?

Let's say I maintain dozen of packages, then I would rather like to write it once and include it everywhere. I could copy and paste all the common functions instead of including, but if I notice a bug or have to upgrade the code for some reason, then I have to edit the code dozen times. But if I only have to maintain one package, then I prefer the copy&paste approach.

For trivial functions, I'd rather just copy them into my source code.

I'd use standard String.prototype.padStart. For anything more complicated, but not too complex (say "libraries" with 100-300 lines of code), it's very tedious and time consuming to find something on npm that satisfies the specific contraints of my app precisely. This certainly fits in that category.

For libraries above 1-2k lines of code that can be quite complicated (eg. things like database connectors), it starts being reasonable to mold expectations of my app around the library.

Also using coherent large utility libraries is something worthwhile (or was in the past), libs like underscore/lodash, etc.

But composing app from random underdocumented 2 function "libraries" from npm is just pain if you know what you're looking for. 1) you can't trust the code, so 2) you have to read it anyway and reading is often slower than writing

Searching for a package on npm is more time-consuming than writing 300 lines of code? Perhaps your app has some very specific constraints, but for a simple library like this, I have a hard time believing you could code something up on the fly faster than it would take to npm install the package in question.

I have absolutely written a diffing algorithm like this faster much than it would take to find an appropriate version on npm.

Every so often I try to reach for an npm library, only to find out that it's suboptimal, doesn't handle my edge cases, or just plain worse than the code I would have written.

The last time I tried to do this, I wanted a hash map with custom equality functions on the keys, since JS doesn't have one built-in by default. I searched for about 2 hours and didn't find a single one that actually worked (several claimed to be, but actually didn't work...)

So I wrote my own in about 100 lines. [0]

Data structures and algorithms tend to be really awful on npm.

[0] https://github.com/magcius/noclip.website/blob/master/src/Ha...

In the long term, having to manage all these dependencies, keeping track of incompatibility, testing updates, or worse, the lack of management, is time consuming.

I agree it can be pretty easy to set up a basic version of this, but eliminating edge cases, optimizing, and unit testing can make it take significantly longer.

Reply on news.ycombinator.com