So let's try to make this thread more continuous. 🤓 #programming So, I made reestablished the static deployment of the web server. ✨ I'm noticing the bad state of my web server more and more, especially caused by hard coded paths. #http #blog Thereby, I also found out, that XSL does not consistently support relative paths. 😞 I'm now solving this via stylesheet parameters, where the Java program determines the relevant absolute paths before executing XSL. #XSL https://github.com/www-splitcells-net/net.splitcells.network/issues/42 https://www.oxygenxml.com/archives/xsl-list/200403/msg01013.html #XSL Starting to make API documentation. 🤓 https://github.com/www-splitcells-net/net.splitcells.network/commit/0fdbf09adf207961a307a0910e260f5fb45ca2dd http://splitcells.net/net/splitcells/dem/guidelines/index.html#d20e340 #documentation Finally, got CI working everywhere again! 👯‍♂️ https://github.com/www-splitcells-net/net.splitcells.network/issues/39 #CI A new search bar concept deployed at the top left -> not pretty but sweet 🍬 http://splitcells.net/ #html #css

So, I have done a little break and am getting back 🤓 The configuration of the repair algorithm is now improved and now I can solve the oral exam sample problem somewhat consistently. Now I need to add the test case to CI and manually validate the solution. https://github.com/www-splitcells-net/net.splitcells.network/issues/7#issuecomment-807652333 #optimization The improvement consists of making each escalation step as small as possible, in order to check the potential of each escalation step. Bigger escalation steps seemed to just cause a performance hit, without improving the result. So making the escalation steps smaller, frees computing resources for precise repair steps. In related news, the server/rendering code of my website is now public. Unfortunately, it is a bit specific to my PC, because XSL does not really seem to support relative paths for file access consistently 😕 https://github.com/www-splitcells-net/net.splitcells.network/issues/14 #web Also, the publicly generated website is not useful yet, because no styling is working. So I still have to rely on my private extensions of the server, in order to generate my public website. In the end, the state of the server is bad, as it is not easily adaptable and not even executable on other computers without some configuration. https://github.com/www-splitcells-net/net.splitcells.network/tree/master/projects/net.splitcells.website.server #web First of all, I need to create a spec, which can be implemented in any way and where each implementation can be combined into one renderer, via a simple shell pipeline. That should make the architecture a lot smaller and adaptable. https://github.com/www-splitcells-net/net.splitcells.network/issues/26 #web

So I found the issue, why the Repair algorithm created an empty list of repair steps: It was due to the fact, that a merge function, did not correctly merge multiple Maps. Instead, it created empty Maps as a result, which itself resulted into missing allocation events. https://github.com/www-splitcells-net/net.splitcells.network/commit/8fb84cbc63dc7df4f55738eb0215dd023463ec24

Now there are no undefined variables after each optimization step left. 🥰 Unfortunately, I found out, that the ConstraintGroupBasedRepair also has sometimes some problems with some duplicate removal events. 😮‍💨

Also, in its current state the ConstraintGroupBasedRepair is not able to solve the oral exam problem. It gets close sometimes and creates as solution with a single digit number of errors, but the optimizer does not solve the problem anymore. -.-

I'm doing small baby steps, one at a time. It will be fine. https://giphy.com/gifs/this-is-fine-QMHoU66sBXqqLqYvGO/links

The see me rollin, they hatin. https://9gag.com/gag/aDdgLzG This weekend I did not work on Gel unfortunately.

So the Java code for the website server is published, but the public website data is not migrated yet. https://github.com/www-splitcells-net/net.splitcells.network/issues/14

Thereby, I noticed how messed up the code is. So I planned the simplification of the code, but it has not the biggest priority: https://github.com/www-splitcells-net/net.splitcells.network/issues/32 #technicaldebt

In order to simplify the software, the software should be also easy replaceable: https://github.com/www-splitcells-net/net.splitcells.network/issues/33 In other words, the website server should have a life cycle, where it is quite possible, that it may be deleted in the future.

So I tried to setup up both of my Raspberry Pies as local build servers. My older one (v2) worked, but booting from USB cannot be made to be working. Instead I attached an USB as an additional external driver, in order to improve its speed and to minimize the tear on the SD card. I hope that works.

As hard it is to believe standard out of the box Java 11, does not run on Raspberry Pi. This example demonstrates that the reparability and maintenance of hardware is a real challenge. In this case, it does not work out of the box, because the floating point arithmetic of the CPU is not supported anymore. Instead one has to use an build of the OpenJDK, which sidesteps this problem. Reparability and maintenance have a cost and should not be underestimated.

I could not get my newer Raspberry Pi (v4) running via USB. I will have to buy a mini HDMI adapter in order to see, what the issue is. We'll see... 🙄

It was an interesting experience, to program in my native language, but it is now time to go back to English. So I translated Gel back from Latvian to English. I only needed to edit 8664 lines of code. Thank god it is done! At least Intellij's auto refactoring helped a lot. https://github.com/www-splitcells-net/net.splitcells.gel.core/commits/master?after=bf68cd0943f0d53da50ce585b740c92443a033c5+34&branch=master

Wait what? This was only the core project? Logging, runtime checks, tests and documentation needs also to be ported? NnOOOOOOOOooooo!!!1! https://www.youtube.com/watch?v=nIAYtHiCjN8

So I suspected, that the first wave of vaccine delivery will be a fucked up (in Germany). 😓 https://twitter.com/splitcells/status/1343279479720968199

It may be the case, that Bavaria is too incompetent in order transport the first wave of vaccines for its people. So it seems, that I remain to be correct. https://www.merkur.de/bayern/corona-bayern-soeder-impfstoff-biontech-huml-campingboxen-panne-news-aktuell-zahlen-zr-90160295.html

Unfortunately, I did not publish my incompetent predictions earlier: there will be lockdowns up to April. After that the lockdowns will probably be lifted (at least mostly) as COVID infections seem to be seasonal. https://www.google.com/search?channel=fs&client=ubuntu&q=Germany+covid+infections

Also keep in mind that the German federal elections are going to happen this year.

So now there is a blog page for this journey. It helps me to right it down before publishing this on other sites: http://splitcells.net/net/splitcells/martins/avots/website//blog.html

The current Gel projects are also available online: https://github.com/www-splitcells-net/net.splitcells.gel.doc/tree/master/src/main/xml/net/splitcells/gel/doc/project http://splitcells.net/net/splitcells/gel/doc/index.html

I have also had the idea in my mind for some time, to abandon JSON and XML for personally create data like text. JSON's inability to model namespaces and schemes in a standard way and XMLs complexity and bad programming integration just causes problems over and over again. So I have made a low priority project out of this, where I will right down my thoughts on this one. https://github.com/www-splitcells-net/net.splitcells.dem/blob/master/src/main/xml/net/splitcells/dem/project/data-as-source-code.xml

I do this especially because I have made some attempts and most of them were 70% okayish but not good enough. If they were organized better like one project, maybe this problem would be solved. In fact I am very sure, that this problem would have been already solved, when I am looking a other similar project. Ggggrrrrr.... Software is hard as fuck I have made so many **** mistakes.... and I'm still probably doing stupid shit as well ^^

I have now integrated my project cluster into Github https://github.com/www-splitcells-net with a simple CI integration. Everything can be build by downloading https://github.com/www-splitcells-net/net.splitcells and executing "bin/build".

A Gitlab integration is also working: https://gitlab.com/splitcells-net . I mainly did this, in order to ensure, that the projects are easily moveable and can be build on different systems.

Unfortunately, some parts of the Gel Project will remain private. Currently only the core of the program itself is published and this may stay this way for some time.

Tests, consistency and runtime checks and examples of project usage, will stay private as there is no advantage in publishing these yet and I am afraid that this won't change this.

This hinders development by third parties, but let be frank to one another: probably no one will be interested in this random solution project (Gel).

It's 2021: if you are doing solution stuff and you are not doing a Kubernetes cluster with tensorflow and GPUs, what are you even doing?

Have the 90s called back? Oh right, my fault: the second world war has called and wants its operation research back: https://en.wikipedia.org/wiki/Operation

"In every crisis, there is an opportunity" somebody Corona is synonym with change. Living in Germany I think, that the first wave of vaccine, will be a little fuck up. #COVIDー19 #Impfstart

Therefore, I have no clue how long this crisis will take place and it's time to take actions. During my time in university I worked on an solution program based on assignment problems.

It's main purpose is to understand assignment problems, apply optimizations and to analyze these. The problem focused is placed on scheduling problems and the field of solution research.

Caused by the project's history, scheduling problems in school are used as test problems in order to test new development https://t.co/kaUuziyNLS?amp=1

After my studies I worked on the project a bit, but unfortunately, I have a chosen a very bad approach in order to develop the next version.

Instead of just working on the software like normal, I kind of created a new version scratch. I did migrate a lot of of code to the new version, but the solution system, the tests, the constraint system were rewritten from scratch.

The architectural changes were sound, but the development should have taken place by migrating the whole project step by step, while ensuring that all functionality is still working

That is kind of a boring way to develop things at first sight, but time after time I learned, that this method is in fact more fun, because after a time you have giant system, where you are adding new and new features, which just work (tm) 😉 https://youtu.be/GmUtltvUQX0?list=PLcXPLd_I-oBpdLobK860iqyjqEE3lxnEg

Aaanny way, after a time I started to make this project public, but with a twist: I developed the program in Latvian in order to try out my native language: https://t.co/Nq2ruKV3Cx?amp=1

I also experimented with streaming the development. Of course nobody watched, but it was a nice psycho test: https://www.twitch.tv/splitcellsnet

So we have a crisis, an opportunity, a failed project and a nice cozy place. https://twitter.com/splitcells/status/1343300718170660867/photo/1

So I'm starting to develop the Generic Allocator (Gel) in English again and I'll eventually start to publish private parts of the project too: https://t.co/eTsY1tbmFJ?amp=1

So starting with this I will stroke the fires of my a ambition for this project and hope for the best. https://twitter.com/splitcells/status/1343307545247686658/photo/1

As a start I published the source code of Gel's documentation: https://t.co/I7wcQtC6Nx?amp=1