Do all the things like ++ or -- rants, post your own rants, comment on others' rants and build your customized dev avatarSign Up
From the creators of devRant, Pipeless lets you power real-time personalized recommendations and activity feeds using a simple APILearn More
Search - "yaml really?"
Dev manager: great news guys. We’ve built a new tool to do automated testing on apps. We’ve gotten rid of the old Appium solution we were using and built this new one.
Me: why not just use the inbuilt native stuff? Click to record works really well.
Manager: nah we thought it would be more flexible to build it ourself.
Me: ... ok ... moving on ... how does it work?
Manager: well this new .jar, you download it, pass in a config file, setup up your simulator and appium and the jar will do everything for you.
Me: ... wait you said you hate Appium? Now you’ve built a wrapper around it? And it doesn’t even set everything up, you’ve to do it all by hand?
Manager: oh we had too, would be too much effort to replace it. Don’t worry we can now write all our tests in .yaml config files instead of using Appium.
Me: so we’ve lost the ability of auto-complete and type ahead, everyone has to upskill on a new tool, it offers no new features over what’s available out of the box and we’ll have to deal with new bugs and maintenance and stuff our self ... because we need more flexibility?
Manager: oh don’t worry. The guy who built it is staying here. He’s going to deal with bug fixes and add features. He’s only one guy, but he’s really sharp, it’ll be great for us and the team.
Me: ... ... ...
*audible noise of soul breaking*
Me: ... ok thank you. I’ll look into this new tool3
love-hate relationship with Python semi-rant
The year is 2020.
I have already grown accustomed to the idea that in order to do ML without worrying too much about having to completely jump through hoops with the tech stack I have chosen that I would have to settle with Python, which I like.....for small scripts that don't do much other than piping data around or doing simple admin tasks, that is generally our use of Python at work.
For anything bigger I would prefer something else. Not because I find anything inherently horrible in Python, I find it to be a nice language overall, that has made it possible for many to find a passion inside of the world of development and possibly an interesting in overall engineering and computer science principles. Much respect Python, good game Guido VR, what you did changed the world.
But it is that damn whitespace that gets me, the need to use it as a way to properly write blocks, I just can't make myself like syntactical whitespace no matter what I do. I can do without static typing, shit I did it for the longest time with JS way tf before Node and Typescript were a thing, and I have done it before PHP's attempt at having type hints, which still leave much to be desired. Ruby(imho) the most elegant language around doesn't have it and that is fine really, it does not bother me as much, if mypy gets powerful and widely adopted enough it will then be a non-issue.
But another thing that the 4 languages i mentioned before have is non-existent syntactical whitespace......I just can't stand it.
So, why am I saying all of this nonsense? Today I wanted to recreate a conda environment and landed on the use of YAML............which has syntactic whitespace and I lost my shit.
I seldom bitch about languages and technologies, shit, I used VBScript before, not only did I get paid handsomely for it, but I fucking enjoyed it(probably cuz I am a masochist).
But two things I cannot abide: VBA and syntactic whitespace.
Once I get enough knowledge for it I will push for the same level of tooling in Python to be ported to Scala.
Thank you for coming to my whiny post about something as small as bitching about syntactic whitespace.9
Background: We switched from just simple old PHP and JS using notepad++ to PHPStorm and its infinite configurables, Symfony 4, Twig, Composer, Doctrine, Yarn, NPM, Bootstrap, ( thank the stars we didn't try to add Docker in with all this ), any other junk I'm missing here? Then upgraded to Symfony 5.
Symfony's autowiring: madness behind the curtains. I get frustrated about when and where I can just magically inject these dependencies or use config variables, you know, like the ones you define in service.yaml. Hmmm, "service".yaml. In a controller you can say getParameter() but in a service you have to inject the parameter, FROM THE "SERVICE".yaml!!! Autowiring drives me nuts. Ok, so we can supply dependencies using the constructor, that's great! Within a controller you never have to instantiate the object you're passing to the constructor (autowiring handles that). That's cool, weird when we you try to trace it for the first few times, but nice I guess. Feels like half-assin' it. What bugs me here is that it only works in controllers... I guess out of the box.. i'm not even sure. To get that feature to work for services you have to make some yaml edits. Right?Maybe? Some of the Symfony tutorials have you code up some junk then trash it. Change config then wipe that out and do X instead... so I have no idea what "out of the box" for Symfony really is.
Found this cool article that describes my frustrations in better terms and seems like a good resource to learn about autowiring. I need to continue my yaml wizardry classes. https://alanstorm.com/symfony-autow...
.....And on to YAMLs, or CSS, or JS or any other friggin' change you make to a file anywhere... Make a change, reload page, nothing... nope you have to do some hidden cheat combo of yarn dostuff -> cache:clear -> cache:warmup -> cache:cache:the:cache ... I really really hate this crap. Maybe I'm too old school for all this junk. It was simple with pure PHP. Edit code, push file, reload page, and oh look it changed! Done. So happy! Ok, Ok. Occasionally the js or css might get cached by the browser and you have to ctrl/f5 or Shift/f5 .. one of those. With this framework there's just so much more that you have to remember to do get some new feature of your site loaded.
Now, I totally get wanting to use some type of entity framework, but I feel like my entire world turned backwards. Designing tables using something like MySQL Workbench made sense. I can see all the columns and datatypes right there as i'm building them. From what I've experienced now with Symfony/Doctrine is you have to make and entity, get a shit-ton of question lobbed at you and if it's a relation field you have to really have a clear idea of the cardinality up front. Then we migrate that to the database. Carefully read through the SQL if you really really just want to use migrations:migrate in Prod. That alter table could cost you some some downtime if your table is large.
Some days man....
So this is more like an accident rather than a dissapointment but the thing goes like this.
I make multivariant stacks with vagrant, shell script loaded as stated in yaml config and every bit and piece done from there the only times it disapointed me was the first time i lost it and when i got asked to make it tomcat compatible, not that the person really needed tomcat but just because i called it a stack initialiser it must have tomcat support.
Some really motivated guy.
He apparently wants to monitore his opensource application on his spare time.
His application is likely to have no users though.
But well, that guy looks like kinda montivated.
For professional purpose, guy already did monitore with newrelic.
Seems like he was not satisfied and switched to datadog 3 years ago.
But liking digging dirt, he migrated to self hosted telegraf/influx/grafana (which he likes to about)
Today that guy is not in his company but on his potatoe machine in the cloud. So he wants to be minimalistic, datadog should do.
Now you got it, random ff*** is me, on a weekend, a shinny saturday for that matter.
Actually now it is night.
Now let's start the fight.
I have datadog scripts!
But datadog be sneaky as well. datadog upgraded to v6 8=)
-> scripts ain't working. outdated.
I check the logs. Too bad!
-> datadog removed dogstatsD.log in v6!
Well I have nothing to do in my life it is too cold outside as they say. I read the (sluggy) datadoc and tries some shell command (given in doc) to upload some events to dogstatsd (via udp).
-> Nothing happens, neither in local nor in remote.
ok maybe command not up to date, so let me try some official library. datadog from python. Feels like a nice try!
-> only available for python >= 3.5. 3.4 on my good ol' jessie. Upgrading os for datadog not acceptable.
Maybe dogstatsD not started... doc says it is by default, but well, not the first time doc is wrong... I put datadog as log verbose. Guess what: as per standard: shitload of error.
Digging... kubexx, docker and whatsoever apparently preventing collector to do its normal stuff
np, I am gonna check that on github! Goog, people have the same errors. They seem to fix it by trying some settings, with. or without luck
-> I am not that warrior to check every stuff
Ok, let's stop the datadog events, it works. It does not anymore. You know that sentence. We all know it.
Still not enough!
How about testing that uber super nice feature of v6. The logs. After all I want to make events out of my applicative logs.
How about reading the log again. Configure the yaml log as they say. Done. Make some pattern. Read the best practive. Done. Configures the yaml. Done. Now testing.
-> remote datadog interface be like: no logs for you dude you need to pay
Fuck datadog, fuck that v6 version, good old tail -Fxx | someaggreate.js|sendmail will do...