Continuing posting things I have made to try and #GetFediHired . This time, it's Covenant: a Turner-total, call-by-push-value, hash-consed IR, implemented in pure #Haskell https://hackage.haskell.org/package/covenant

This was a _very_ challenging project. I not only helped win the funding that made it happen, I also had to solve multiple theoretical and practical problems while doing so. It was however, very fun, and I even had Paul Levy drop in once to meet the team!

As I am finishing up at $CURRENT_JOB, I made a small #Haskell library designed to solve a problem we encountered with QuickCheck, specifically shrinkers. Since I'm looking to #GetFediHired, I figured I'd share it.

Hackage: https://hackage.haskell.org/package/quickcheck-helpers

Github: https://github.com/mlabs-haskell/quickcheck-helpers

To make it easier to #GetFediHired , I have updated my site to include my skills and experience: https://retro-freedom.nz/about/

This is not necessarily absolutely everything I have ever done, and I may add more to it, but it's a good sampling. I tried to focus on projects other people can verify, rather than proprietary work I cannot really demonstrate.

I am looking for remote opportunities. Short or long-term is fine. I am New Zealand-based.

Continuing my daily posting of things I've made to try and #GetFediHired, this is a contribution I made to the `bytestring` library to give them a SIMD-optimized UTF-8 validator: https://github.com/haskell/bytestring/blob/master/cbits/is-valid-utf8.c

This was, without a doubt, some of the most challenging code I have ever written in _any_ language. It involved a very complex mixture of concerns, and I had performance targets to hit with it too. Oh, _and_ it had to be self-contained and portable, including to 'no SIMD' platforms!

In a further bid to #GetFediHired, I'm going to post some of the work I have done. I'll post something every day!

Today, it's a blog post talking about different string matching implementations in Haskell, how we can assess their performance, and a reasonably clear explanation of the BNDM algorithm and the source of its speed. It's on my soon-to-be-former employer's blog: https://www.mlabs.city/blog/first-know-thy-self-understanding-string-matching-performance-with-ascii

(my name is at the _very_ bottom, in small font)

#Haskell #Algorithms

So, I just got my four weeks' notice. Thus, I am looking for new remote opportunities, short-term, long-term, whatever term really.

I'm based in New Zealand, and I've worked with a mixture of Haskell and C professionally for about six years. Some of the things I've built:

* Several eDSLs
* Optimizations for various interpreters
* `bytestring`'s UTF-8 validator with SIMD acceleration

I've also applied for, and received, multiple grants. Am able to learn fast too!

#GetFediHired

These are cattails. https://www.youtube.com/watch?v=RWEk_Gp85w0

More seriously, this is becoming a marathon #Terrain project.

It has been 0 days since I wasted both time and money on a paid #FDM design that didn't work or had severe deficiencies I had to fix myself.

Ugh.

'4' has become the '67' of tech, and I am not happy about it.

TIL about the Bernays-Schonfinkel class. What I didn't learn is what Moses Schonfinkel had to do with it - I know of him only by way of the combinator calculus.

Any #logic people able to enlighten me?

In my #LaserEngraving adventures, I finally cracked the code to get some proper depth on 3mm acrylic engravings without getting a warped potato chip.

It comes down to the following:

* Low-energy fast passes (I'm doing 9000mm/min at 30%)
* Cross-hatch the engraving
* Lots of passes (picture is six)
* The _gentlest_ bit of air assist

Really happy with that - decent depth and no warping!

Any #3DModelling Fedi friends want to make some money? I'd like a model of some cute animals, as a relief engraving, suitable for #CNCMachining, and am happy to pay for it. This must be your own work - I have a zero-tolerance policy for slop.

Reach out to me in a PM if you're interested.

Dear image-expert Fedi: I would like to do some relief engravings, and would love to know of any depth mapping tools (preferably open source) you can recommend. The target is a CNC engraving, but since I have a laser, that would be cool too.

I don't mind doing some learning, but I genuinely have no idea how to get good depth maps for either process from a photo.

#CNCMachining #LaserEngraving

Today, we used #CNCMachining, #LaserCutting _and_ #LaserEngraving . This will eventually be a DnD dicebox, but the lid will involve a paint fill, so I won't finish it today.

Those who follow me for #FDM , fear not: this is a swingy lid box, so it'll need a (printed) latch.

Lastly, set each of the ends of each perpendicular line _not_ on the chords to be on the other perpendicular line. This will combine into one point, which is your centre.

I find this is often useful when you want to either widen, or shrink, a hole, while keeping its centre in the same place.

Adapted from here: https://www.instructables.com/How-to-Find-the-Center-of-a-Circle-1/

5/5

Then, construct a perpendicular line from each of the centre points on each chord. No need to worry how far they extend for now.

4/?

Then, put points in the middle of each chord. FreeCAD can do this automagically by doing a symmetry constraint on the point using the endpoints.

3/?

Next, connect these dots using chords. Set them to construction geometry.

2/?

One common thing I end up doing in #FreeCAD is fixing up (frankly) poorly-designed models I find online. Since these are usually STLs, it leads to some annoying moments, like 'finding the centre of a circle that's now actually a bunch of vaguely-circular line segments now'.

Decided to showcase an easy technique in part design to find the centre of such a 'circle'. First, mark three points on the perimeter - doesn't matter where.

#CAD

Following up from my baking soda #Laser experiments, I made this engraving for a friend. The material was slightly warped, which is why the tip of one claw didn't fully engrave, but otherwise, a pretty great result.

25% power was indeed the right call here. Cutting through the 5mm material took a little bit of effort though.

#LaserEngraving #LaserCutting