Graduated 2020 ; PhD thesis

Graduated in 2023; PhD thesis.

I was expecting some references to the literature on clone detection and clone management to support the motivation for something better than a posteriori diagnosis and treatment.

Paul is the C++ expert at Philips who supported making the transformations and testing the results.

I'm one of the co-designers and maintainers of Rascal and helped from a distance with the transformations, and co-authored the paper.

Rodin is the author of Clair, the C++ front-end for Rascal based on the Eclipse CDT; this enabled the entire process. He helped writing the scripts and is a main author of the paper with Mathijs.

Mathijs did the maintenance work, writing the Rascal scripts and grammars, testing the script, testing the output, and making sure Rascal was accepted in the Philips software process.

https://arxiv.org/abs/2111.05132

https://www.labri.fr/perso/falleri/perso/

https://tdegueul.github.io/

https://lmove.github.io/

public or private comments on this talk are very welcome.

dank voor deze suggestie.

working now on a programming language with explicit automated accuracy estimates; with distributive law between * and +, as it should be :-) Contact me at Jurgen.Vinju@cwi.nl if you are interested!

This is a small group of people. Even programming language experts are not among them; they take hardware (often) for granted. Even the functional programming language Haskell, which is supposed to improve program correctness by construction, features floating point numbers.

this makes floating point errors in to an "unknown unknown" since even for a critical observer there is no practical frame of reference other than doing the computations on paper, manually.

this link is broken

Is average time a good metric if you have long tail distributions? Why not median? Or is it already median? Then I'm in trouble and its good to see this.

that's spam

Graduated in 2019

Graduated in 2019; Duo PhD thesis

https://en.wikipedia.org/wiki/Computer_literacy

https://en.wikipedia.org/wiki/Digital_literacy

I learned this from Magiel Bruntink; in medical research we require baseline statistics to describe fluctuations of measurement points in the background of the new invention under study. If normally people heal within 14 days anyway by themselves, the result that all of them healed within 20 days sounds wonderful but means nothing. How many results in software engineering have such wonderful data without the background variability even mentioned?

I'm often flustered when "evaluation" of a new analysis tool is called "validation" or when the validation of an evaluation research method is called "evaluation" instead in academic literature. My trick to remember is that the tool is required/claimed to have "value" and this it needs to be evaluated to bring this to light or to otherwise. The evaluation however is required/claimed to be trutful/valid. Hence it requires at least some external or orthogonal validation next to the common knee drops in threats to validity sections.

trying to control for the common situation where arbitrary code is written without any solid analysis I've used "professional" here. I admit it's a weak solution.

this is a creative interpretation by my hand, which I like :-) Definitions of the field of software evolution vary greatly and confuse me a lot.

https://nl.wikipedia.org/wiki/Ariane_5

software which is only working accidentally, i.e. there is no known reason for this, or this reason has been lost. this is an all too common situation where responsibility is suddenly out of the equation and there are only victims.

a purposeful wishful statement, if not imperative.

subtitle: they simulate Church's and Turing's models of automatic computation up to the practical memory and time limitations, for which outside of quantum computing theory, there exists no known generalization.

https://youtu.be/NfcD3RiB1EA "bellen? even tot 10 tellen" is another instance of the same that I found this youtube movie on. The original slogan in Dutch is "wie tot 10 kan tellen, kan de hele wereld bellen".

https://en.wikipedia.org/wiki/Volkswagen_emissions_scandal

Graduated in 2018; PhD thesis

This is a color deprived rendering of the original by Vik Muniz. Please check out his art at http://vikmuniz.net/ and go to one of the musea that display his wonderful art.

Unfortunately in the Dutch original writeup I called this an "Escher triangle" because I learned of Penrose triangles via the art of Escher. Escher learned from Penrose so I corrected the mistake here.

Kees van der Laan [1943-2015] of the NTG, the Dutch TeX users group in 1988. Sadly, I've never met him personally but his Postscript libraries and art have made a lasting impression.

https://en.wikipedia.org/wiki/Fermat%27s_Last_Theorem

https://en.wikipedia.org/wiki/Pythagorean_theorem

"information" in a broad sense, which includes manual analysis backed by some theoretical frame that the analyst is using implicitly. This can go horribly wrong if the analyst has the wrong frame. Say a novice analyzer reads a + b, thinks + is the operator they learned about in school and concludes that it is equal to b + a. In floating point arithmetic this is not the case, a different theoretical frame, which leads to radically different conclusions.

https://en.wikipedia.org/wiki/Emergence#Emergent_properties_and_processes

"I just called to say I love you, I just called to say how much I care"

https://github.com/hypothesis/

example tag: <link rel="canonical" href="http://where/your/old/page/is.html" />

Graduated 2011; PhD thesis

Graduated in 2017; PhD thesis

Graduated in 2017; PhD thesis

I'm referring to the way programmers are depicted in the media (both fiction and non-fiction) as "wizkids", "hackers", "nerds", to all the "zeroes and ones", the subliminal references to The Matrix when software is in the news, etc. M

GDPR compliance is at heart a property of the source code that companies and other (government) institutions depend on. I wonder how many eyes have actually inspected the source code that is written about today (indirectly via the business front) as being compliant.

I believe that once two sources of information are selected it is probably "easy" to make their value complement each other. The primary challenge is to select the most valuable pairs, and to be able to cater for as many pairs as possible. All of these context-sensitive software analysis tools pose a design and maintenance nightmare in themselves...

let alone studies which try to compare the process in which people organize themselves to construct software; even more variables are at play that, at least I, have no inkling how to frame.

What Tudor Gîrba always says in his presentations about software inspection and visualisation tools: we should avoid to read code, since it is practically always a futile endeavour.

The field of Software Engineering is about analysis of Software and the way it is designed, constructed and maintained.

The analysis perspective brings the academic level more the foreground and harder to underestimate. Making a baby can be quite simple, but understanding the baby is nearly impossible. Building a LEGO tower is childsplay, while predicting when the tower will topple requires advanced physics and mathematics.

Analysis, deconstruction and the understanding that comes from it, in general is simply a harder task than creation and construction. Great construction involves deep analysis, because it makes the result great by design.

how often have you heard a fellow researcher in Computer Science or Physics say "but that's just engineering", "you only have to code it now that we understand the mathematics", "can you please hack this out for me?".

See also What is code and what is coding?

In particular it would be cool if journalists and movie makers would read this blog. They have a big influence on the perception of the public regarding programmers and software, and what they do.