Now, instead of shouting, I can just type "fuck"

The Fuck is a magnificent app that corrects errors in previous console commands.

inspired by a @liamosaur tweet
https://twitter.com/liamosaur/...

Some gems:

➜ apt-get install vim
E: Could not open lock file /var/lib/dpkg/lock - open (13: Permission denied)
E: Unable to lock the administration directory (/var/lib/dpkg/), are you root?

➜ fuck
sudo apt-get install vim [enter/↑/↓/ctrl+c]
[sudo] password for nvbn:
Reading package lists... Done
...
➜ git push
fatal: The current branch master has no upstream branch.
To push the current branch and set the remote as upstream, use

git push --set-upstream origin master

➜ fuck
git push --set-upstream origin master [enter/↑/↓/ctrl+c]
Counting objects: 9, done.
...
➜ puthon
No command 'puthon' found, did you mean:
Command 'python' from package 'python-minimal' (main)
Command 'python' from package 'python3' (main)
zsh: command not found: puthon

➜ fuck
python [enter/↑/↓/ctrl+c]
Python 3.4.2 (default, Oct 8 2014, 13:08:17)
...
➜ git brnch
git: 'brnch' is not a git command. See 'git --help'.

Did you mean this?
branch

➜ fuck
git branch [enter/↑/↓/ctrl+c]
* master
➜ lein rpl
'rpl' is not a task. See 'lein help'.

Did you mean this?
repl

➜ fuck
lein repl [enter/↑/↓/ctrl+c]
nREPL server started on port 54848 on host 127.0.0.1 - nrepl://127.0.0.1:54848
REPL-y 0.3.1
...

Get fuckked at
https://github.com/nvbn/thefuck

EoS1: This is the continuation of my previous rant, "The Ballad of The Six Witchers and The Undocumented Java Tool". Catch the first part here: https://devrant.com/rants/5009817/...

The Undocumented Java Tool, created by Those Who Came Before to fight the great battles of the past, is a swift beast. It reaches systems unknown and impacts many processes, unbeknownst even to said processes' masters. All from within it's lair, a foggy Windows Server swamp of moldy data streams and boggy flows.

One of The Six Witchers, the Wild One, scouted ahead to map the input and output data streams of the Unmapped Data Swamp. Accompanied only by his animal familiars, NetCat and WireShark.

Two others, bold and adventurous, raised their decompiling blades against the Undocumented Java Tool beast itself, to uncover it's data processing secrets.

Another of the witchers, of dark complexion and smooth speak, followed the data upstream to find where the fuck the limited excel sheets that feeds The Beast comes from, since it's handlers only know that "every other day a new one appears on this shared active directory location". WTF do people often have NPC-levels of unawareness about their own fucking jobs?!?!

The other witchers left to tend to the Burn-Rate Bonfire, for The Sprint is dark and full of terrors, and some bigwigs always manage to shoehorn their whims/unrelated stories into a otherwise lean sprint.

At the dawn of the new year, the witchers reconvened. "The Beast breathes a currency conversion API" - said The Wild One - "And it's claws and fangs strike mostly at two independent JIRA clusters, sometimes upserting issues. It uses a company-deprecated API to send emails. We're in deep shit."

"I've found The Source of Fucking Excel Sheets" - said the smooth witcher - "It is The Temple of Cash-Flow, where the priests weave the Tapestry of Transactions. Our Fucking Excel Sheets are but a snapshot of the latest updates on the balance of some billing accounts. I spoke with one of the priestesses, and she told me that The Oracle (DB) would be able to provide us with The Data directly, if we were to learn the way of the ODBC and the Query"

"We stroke at the beast" - said the bold and adventurous witchers, now deserving of the bragging rights to be called The Butchers of Jarfile - "It is actually fewer than twenty classes and modules. Most are API-drivers. And less than 40% of the code is ever even fucking used! We found fucking JIRA API tokens and URIs hard-coded. And it is all synchronous and monolithic - no wonder it takes almost 20 hours to run a single fucking excel sheet".

Together, the witchers figured out that each new billing account were morphed by The Beast into a new JIRA issue, if none was open yet for it. Transactions were used to update the outstanding balance on the issues regarding the billing accounts. The currency conversion API was used too often, and it's purpose was only to give a rough estimate of the total balance in each Jira issue in USD, since each issue could have transactions in several currencies. The Beast would consume the Excel sheet, do some cryptic transformations on it, and for each resulting line access the currency API and upsert a JIRA issue. The secrets of those transformations were still hidden from the witchers. When and why would The Beast send emails, was still a mistery.

As the Witchers Council approached an end and all were armed with knowledge and information, they decided on the next steps.
The Wild Witcher, known in every tavern in the land and by the sea, would create a connector to The Red Port of Redis, where every currency conversion is already updated by other processes and can be quickly retrieved inside the VPC. The Greenhorn Witcher is to follow him and build an offline process to update balances in JIRA issues.
The Butchers of Jarfile were to build The Juggler, an automation that should be able to receive a parquet file with an insertion plan and asynchronously update the JIRA API with scores of concurrent requests.
The Smooth Witcher, proud of his new lead, was to build The Oracle Watch, an order that would guard the Oracle (DB) at the Temple of Cash-Flow and report every qualifying transaction to parquet files in AWS S3. The Data would then be pushed to cross The Event Bridge into The Cluster of Sparks and Storms.
This Witcher Who Writes is to ride the Elephant of Hadoop into The Cluster of Sparks an Storms, to weave the signs of Map and Reduce and with speed and precision transform The Data into The Insertion Plan.
However, how exactly is The Data to be transformed is not yet known.

Will the Witchers be able to build The Data's New Path? Will they figure out the mysterious transformation? Will they discover the Undocumented Java Tool's secrets on notifying customers and aggregating data?

This story is still afoot. Only the future will tell, and I will keep you posted.

I used to think that I had matured. That I should stop letting my emotions get the better of me. Turns out there's only so much one can bottle up before it snaps.

Allow me to introduce you folks to this wonderful piece of software: PaddleOCR (https://github.com/PaddlePaddle/...). At this time I'll gladly take any free OCR library that isn't Tesseract. I saw the thing, thought: "Heh. 3 lines quick start. Cool.", and the accuracy is decent. I thought it was a treasure trove that I could shill to other people. That was before I found out how shit of a package it is.

First test, I found out that logging is enabled by default. Sure, logging is good. But I was already rocking my own logger, and I wanted it to shut the fuck up about its log because it was noise to the stuffs I actually wanted to log. Could not intercept its logging events, and somehow just importing it set the global logging level from INFO to DEBUG. Maybe it's Python's quirk, who knows. Check the source code, ah, the constructors gaves `show_log` arg to control logging. The fuck? Why? Why not let the user opt into your logs? Why is the logging on by default?

But sure, it's just logging. Surely, no big deal. SURELY, it's got decent documentation that is easily searchable. Oh, oh sweet summer child, there ain't. Docs are just some loosely bundled together Markdowns chucked into /doc. Hey, docs at least. Surely, surely there's something somewhere about all the args to the OCRer constructor somewhere. NOPE! Turns out, all the args, you gotta reference its `--help` switch on the command line. And like all "good" software from academia, unless you're part of academia, it's obtuse as fuck. Fine, fuck it, back to /doc, and it took me 10 minutes of rummaging to find the correct Markdown file that describes the params. And good-fucking-luck to you trying to translate all them command line args into Python constructor params.

"But PTH, you're overreacting!". No, fuck you, I'm not. Guess whose code broke today because of a 4th number version bump. Yes, you are reading correctly: My code broke, because of a 4th number version bump, from 2.6.0.1, to 2.6.0.2, introducing a breaking change. Why? Because apparently, upstream decided to nest the OCR result in another layer. Fuck knows why. They did change the doc. Guess what they didn't do. PROVIDING, A DAMN, RELEASE NOTE. Checked their repo, checked their tags, nothing marking any releases from the 3rd number. All releases goes straight to PyPI, quietly, silently, like a moron. And bless you if you tell me "Well you should have reviewed the docs". If you do that for your project, for all of your dependencies, my condolences.

Could I just fix it? Yes. Without ranting? Yes. But for fuck sake if you're writing software for a wide audience you're kinda expected to be even more sane in your software's structure and release conventions. Not this. And note: The people writing this, aren't random people without coding expertise. But man they feel like they are.

security fiasco due to a malicious npm package:

Because of a bitcoin miner present in event-stream npm module (https://bleepingcomputer.com/news/...), my entire team and I had to scan all our nodejs apps, repos and the most excruciating one, all node_modules folders across all our dev machines and servers, to see if event-stream and flatmap-stream is present, then not just delete it but update a bu**load of upstream dependencies which internally used event-stream. All due to one malicious package which was hidden several layers beneath.

And, this happened almost 8 months after the aforesaid vulnerability was first found.

Every layout goal must take hours of frustrating intuition-destroying trial and error, followed by documentation cross-examination, MRE building, upstream bug-filing, and workaround pursuits.

https://jsfiddle.net/uz5dr8h4/21/

But no, CSS doesn't suck, you're just bad at it.

I had the idea that part of the problem of NN and ML research is we all use the same standard loss and nonlinear functions. In theory most NN architectures are universal aproximators. But theres a big gap between symbolic and numeric computation.

But some of our bigger leaps in improvement weren't just from new architectures, but entire new approaches to how data is transformed, and how we calculate loss, for example KL divergence.

And it occured to me all we really need is training/test/validation data and with the right approach we can let the system discover the architecture (been done before), but also the nonlinear and loss functions itself, and see what pops out the other side as a result.

If a network can instrument its own code as it were, maybe it'd find new and useful nonlinear functions and losses. Networks wouldn't just specificy a conv layer here, or a maxpool there, but derive implementations of these all on their own.

More importantly with a little pruning, we could even use successful examples for bootstrapping smaller more efficient algorithms, all within the graph itself, and use genetic algorithms to mix and match nodes at training time to discover what works or doesn't, or do training, testing, and validation in batches, to anneal a network in the correct direction.

By generating variations of successful nodes and graphs, and using substitution, we can use comparison to minimize error (for some measure of error over accuracy and precision), and select the best graph variations, without strictly having to do much point mutation within any given node, minimizing deleterious effects, sort of like how gene expression leads to unexpected but fitness-improving results for an entire organism, while point-mutations typically cause disease.

It might seem like this wouldn't work out the gate, just on the basis of intuition, but I think the benefit of working through node substitutions or entire subgraph substitution, is that we can check test/validation loss before training is even complete.

If we train a network to specify a known loss, we can even have that evaluate the networks themselves, and run variations on our network loss node to find better losses during training time, and at some point let nodes refer to these same loss calculation graphs, within themselves, switching between them dynamically..via variation and substitution.

I could even invision probabilistic lists of jump addresses, or mappings of value ranges to jump addresses, or having await() style opcodes on some nodes that upon being encountered, queue-up ticks from upstream nodes whose calculations the await()ed node relies on, to do things like emergent convolution.

I've written all the classes and started on the interpreter itself, just a few things that need fleshed out now.

Heres my shitty little partial sketch of the opcodes and ideas.
https://pastebin.com/5yDTaApS

I think I'll teach it to do convolution, color recognition, maybe try mnist, or teach it step by step how to do sequence masking and prediction, dunno yet.

https://i.imgflip.com/2i02zy.jpg

git branch -r
origin/204/match-dsteem-on-sign-transaction
origin/305-support-hive-legacy-api
origin/307-call-async
origin/72-http-socket-support
origin/HEAD -> origin/dev
origin/appbase-http
origin/chore/fix-ws
origin/default-server

but

git push --follow-tags https://github.com/lopudesigns/... --set-upstream origin dev
fatal: refs/remotes/origin/HEAD cannot be resolved to branch.

wut

f***ed with Nginx + php7.0 FPM

connect() to unix:/run/php/php7.0-fpm.sock failed (11: Resource temporarily unavailable) while connecting to upstream

#socket_vs_tcpip