Oh, man, I just realized I haven't ranted one of my best stories on here!

So, here goes!

A few years back the company I work for was contacted by an older client regarding a new project.

The guy was now pitching to build the website for the Parliament of another country (not gonna name it, NDAs and stuff), and was planning on outsourcing the development, as he had no team and he was only aiming on taking care of the client service/project management side of the project.

Out of principle (and also to preserve our mental integrity), we have purposely avoided working with government bodies of any kind, in any country, but he was a friend of our CEO and pleaded until we singed on board.

Now, the project itself was way bigger than we expected, as the wanted more of an internal CRM, centralized document archive, event management, internal planning, multiple interfaced, role based access restricted monster of an administration interface, complete with regular user website, also packed with all kind of features, dashboards and so on.

Long story short, a lot bigger than what we were expecting based on the initial brief.

The development period was hell. New features were coming in on a weekly basis. Already implemented functionality was constantly being changed or redefined. No requests we ever made about clarifications and/or materials or information were ever answered on time.

They also somehow bullied the guy that brought us the project into also including the data migration from the old website into the new one we were building and we somehow ended up having to extract meaningful, formatted, sanitized content parsing static HTML files and connecting them to download-able files (almost every page in the old website had files available to download) we needed to also include in a sane way.

Now, don't think the files were simple URL paths we can trace to a folder/file path, oh no!!! The links were some form of hash combination that had to be exploded and tested against some king of database relationship tables that only had hashed indexes relating to other tables, that also only had hashed indexes relating to some other tables that kept a database of the website pages HTML file naming. So what we had to do is identify the files based on a combination of hashed indexes and re-hashed HTML file names that in the end would give us a filename for a real file that we had to then search for inside a list of over 20 folders not related to one another.

So we did this. Created a script that processed the hell out of over 10000 HTML files, database entries and files and re-indexed and re-named all this shit into a meaningful database of sane data and well organized files.

So, with this we were nearing the finish line for the project, which by now exceeded the estimated time by over to times.

We test everything, retest it all again for good measure, pack everything up for deployment, simulate on a staging environment, give the final client access to the staging version, get them to accept that all requirements are met, finish writing the documentation for the codebase, write detailed deployment procedure, include some automation and testing tools also for good measure, recommend production setup, hardware specs, software versions, server side optimization like caching, load balancing and all that we could think would ever be useful, all with more documentation and instructions.

As the project was built on PHP/MySQL (as requested), we recommended a Linux environment for production. Oh, I forgot to tell you that over the development period they kept asking us to also include steps for Windows procedures along with our regular documentation. Was a bit strange, but we added it in there just so we can finish and close the damn project.

So, we send them all the above and go get drunk as fuck in celebration of getting rid of them once and for all...

Next day: hung over, I get to the office, open my laptop and see on new email. I only had the one new mail, so I open it to see what it's about.

Lo and behold! The fuckers over in the other country that called themselves "IT guys", and were the ones making all the changes and additions to our requirements, were not capable enough to follow step by step instructions in order to deploy the project on their servers!!!

[Continues in the comments]

parsing html with regex, thats something for /r/Whatcouldgowrong/

I am much too tired to go into details, probably because I left the office at 11:15pm, but I finally finished a feature. It doesn't even sound like a particularly large or complicated feature. It sounds like a simple, 1-2 day feature until you look at it closely.

It took me an entire fucking week. and all the while I was coaching a junior dev who had just picked up Rails and was building something very similar.

It's the model, controller, and UI for creating a parent object along with 0-n child objects, with default children suggestions, a fancy ui including the ability to dynamically add/remove children via buttons. and have the entire happy family save nicely and atomically on the backend. Plus a detailed-but-simple listing for non-technicals including some absolutely nontrivial css acrobatics.

After getting about 90% of everything built and working and beautiful, I learned that Rails does quite a bit of this for you, through `accepts_nested_params_for :collection`. But that requires very specific form input namespacing, and building that out correctly is flipping difficult. It's not like I could find good examples anywhere, either. I looked for hours. I finally found a rails tutorial vide linked from a comment on a SO answer from five years ago, and mashed its oversimplified and dated examples with the newer documentation, and worked around the issues that of course arose from that disasterous paring.

like.
I needed to store a template of the child object markup somewhere, yeah? The video had me trying to store all of the markup in a `data-fields=" "` attrib. wth? I tried storing it as a string and injecting it into javascript, but that didn't work either. parsing errors! yay! good job, you two.

So I ended up storing the markup (rendered from a rails partial) in an html comment of all things, and pulling the markup out of the comment and gsubbing its IDs on document load. This has the annoying effect of preventing me from using html comments in that partial (not that i really use them anyway, but.)

Just.
Every step of the way on building this was another mountain climb.
* singular vs plural naming and routing, and named routes. and dealing with issues arising from existing incorrect pluralization.
* reverse polymorphic relation (child -> x parent)
* The testing suite is incompatible with the new rails6. There is no fix. None. I checked. Nope. Not happening.
* Rails6 randomly and constantly crashes and/or caches random things (including arbitrary code changes) in development mode (and only development mode) when working with multiple databases.
* nested form builders
* styling a fucking checkbox
* Making that checkbox (rather, its label and container div) into a sexy animated slider
* passing data and locals to and between partials
* misleading documentation
* building the partials to be self-contained and reusable
* coercing form builders into namespacing nested html inputs the way Rails expects
* input namespacing redux, now with nested form builders too!
* Figuring out how to generate markup for an empty child when I'm no longer rendering the children myself
* Figuring out where the fuck to put the blank child template markup so it's accessible, has the right namespacing, and is not submitted with everything else
* Figuring out how the fuck to read an html comment with JS
* nested strong params
* nested strong params
* nested fucking strong params
* caching parsed children's data on parent when the whole thing is bloody atomic.
* Converting datetimes from/to milliseconds on save/load
* CSS and bootstrap collisions
* CSS and bootstrap stupidity
* Reinventing the entire multi-child / nested params / atomic creating/updating/deleting feature on my own before discovering Rails can do that for you.

Just.
I am so glad it's working.
I don't even feel relieved. I just feel exhausted.

But it's done.
finally.

and it's done well. It's all self-contained and reusable, it's easy to read, has separate styling and reusable partials, etc. It's a two line copy/paste drop-in for any other model that needs it. Two lines and it just works, and even tells you if you screwed up.

I'm incredibly proud of everything that went into this.
But mostly I'm just incredibly tired.

Time for some well-deserved sleep.

Coders cry...

... when they inherit a project that's parsing HTML using regular expressions.

<br>

[CMS Of Doom™]
Ah, yes, their built-in bullshit newsletter module just sent the n-th user n emails. Wonderful considering n=368.

The culprit? Better don't ask...
OK, anyway: So the mailer is running as a CRONjob, but nah, not as a console script call but by a public HTTP GET URL call, fucking obviously (it's the CMS Of Doom for a reason).
So these fucking imbeciles "implemented" an ob_start() callback where HTML links are - for whatever fucking reason - modified by some regex (obviously everybody knows parsing HTML by Regex is trivial). In this case the link was somehow modified to recall the mailer Cronjob...
This must have upset the pngoing mailing process thus spamming mails. Whyyyy

And I've thought I've seen it all after 6 months in this legacy hell...
This is why you don't run a company consisting of only beginners in PHP (in cluding their "CEO")!

So, I've had a personal project going for a couple of years now. It's one of those "I think this could be the billion-dollar idea" things. But I suffer from the typical "it's not PERFECT, so let's start again!" mentality, and the "hmm, I'm not sure I like that technology choice, so let's start again!" mentality.

Or, at least, I DID until 3-4 months ago.

I made the decision that I was going to charge ahead with it even if I started having second thoughts along the way. But, at the same time, I made the decision that I was going to rely on as little external technology as possible. Simplicity was going to be the key guiding light and if I couldn't truly justify bringing a given technology into the mix, it'd stay out.

That means that when I built the front end, I would go with plain HTML/CSS/JS... you know, just like I did 20+ years ago... and when I built the back end, I'd minimize the libraries I used as much as possible (though I allowed myself a bit more flexibility on the back end because that seems to be where there's less issues generally). Similarly, any choice I made I wanted to have little to no additional tooling required.

So, given this is a webapp with a Node back-end, I had some decisions to make.

On the back end, I decided to go with Express. Previously, I had written all the server code myself from "first principles", so I effectively built my own version of Express in other words. And you know what? It worked fine! It wasn't particularly hard, the code wasn't especially bad, and it worked. So, I considered re-using that code from the previous iteration, but I ultimately decided that Express brings enough value - more specifically all the middleware available for it - to justify going with it. I also stuck with NeDB for my data storage needs since that was aces all along (though I did switch to nedb-promises instead of writing my own async/await wrapper around it as I had previously done).

What I DIDN'T do though is go with TypeScript. In previous versions, I had. And, hey, it worked fine. TS of course brings some value, but having to have a compile step in it goes against my "as little additional tooling as possible" mantra, and the value it brings I find to be dubious when there's just one developer. As it stands, my "tooling" amounts to a few very simple JS scripts run with NPM. It's very simple, and that was my big goal: simplicity.

On the front end, I of course had to choose a framework first. React is fine, Angular is horrid, Vue, Svelte, others are okay. But I didn't want to bother with any of that because I dislike the level of abstraction they bring. But I also didn't want to be building my own widget library. I've done that before and it takes a lot of time and effort to do it well. So, after looking at many different options, I settled on Webix. I'm a fan of that library because it has a JS-centric approach. There's no JSX-like intermediate format, no build step involved, it's just straight, simple JS, and it's powerful and looks pretty good. Perfect for my needs. For one specific capability I did allow myself to bring in AnimeJS and ThreeJS. That's it though, no other dependencies (well, at first, I was using Axios because it was comfortable, but I've since migrated to plain old fetch). And no Webpack, no bundling at all, in fact. I dynamically load resources, which effectively is code-splitting, and I have some NPM scripts to do minification for a production build, but otherwise the code that runs in the browser is what I actually wrote, unlike using a framework.

So, what's the point of this whole rant?

The point is that I've made more progress in these last few months than I did the previous several years, and the experience has been SO much better!

All the tools and dependencies we tend to use these days, by and large, I think get in the way. Oh, to be sure, they have their own benefits, I'm not denying that... but I'm not at all convinced those benefits outweighs the time lost configuring this tool or that, fixing breakages caused by dependency updates, dealing with obtuse errors spit out by code I didn't write, going from the code in the browser to the actual source code to get anywhere when debugging, parsing crappy documentation, and just generally having the project be so much more complex and difficult to reason about. It's cognitive overload.

I've been doing this professionaly for a LONG time, I've seen so many fads come and go. The one thing I think we've lost along the way is the idea that simplicity leads to the best outcomes, and simplicity doesn't automatically mean you write less code, doesn't mean you cede responsibility for various things to third parties. Those things aren't automatically bad, but they CAN be, and I think more than we realize. We get wrapped up in "what everyone else is doing", we don't stop to question the "best practices", we just blindly follow.

I'm done with that, and my project is better for it!

Which misanthropic, terrible, perverse excuse for a dogfucker decided that damned non breaking spaces (SPACES!) return false on isWhitespace? It's in the name, space, it's white, it's a fucking white space, a whitespace if you will so who do I have to kill for wasting two damned hours of my life trying to parse away those bastards?

Teach data structures by showing how they're used in real life situations. Don't make us do some nonsense puzzle shit. For example, a friend of mine is learning stacks/queues right now and his assignment is to build a simple HTML parsing algorithm to determine whether an HTML file is valid. This shows the student a practical use of the data structure and reinforces that this shit actually does get used in real life.

Parsing my college's terrible classless html float left div mosaic timetables so I could hide lectures I wasn't taking.

Hey, wanna hear a joke?

Parsing HTML with regex.

I wish there’s a better engine for parsing HTML for relevant content from a URL than mercury for free.

I was making a class project in react and that is not getting it done to my thing. 😑😑

I've to say that javascript is no language I like because I'm more fastinated of building a nice and scalable backend then building a gui (in the browser).

The funny thing is that right now javascript helps me at my current project because many websites implement wide-opened apis for their js frontend and it just works like a charm to use them instead of parsing the whole html and do some XPath stuff to fetch information.

Once upon a time I was working with an engineer who loved sed and awk a bit too much. We had data stored in SharePoint that was retrievable via an RSS feed. Said engineer insisted on using curl to grab the feed and sed/awk to parse the HTML ...

I on the other hand suggested using libcurl (primarily for NTLM auth support) and parsing the RSS feed using libxml.

Which engineer do you think management decided supporting?

Hint: Reusability and maintainability were big requirements in this project.