ttarkowski / quile Goto Github PK

Dear author,

Generally speaking, the library works as described in the readme file of the repo, and the examples can be set up by following the instruction. However, there are minor fixes that can improve the quality of the submission:

• gnuplot is needed to run example1 and generate the output. The instruction for building/running the example can be briefly mentioned in a readme file (similar to mentioning the requirement of BGL for the mithril example). The same thing applies to example 2.
• The mithril example compiles and runs, but there is no output written to the file evolution.dat after 15 minutes. Maybe you need to mention any estimation of the time needed.
• The evanstar example is difficult to evaluate. I think some sample output showing the results the user will get in the end can help to understand what this example tries to do (like the ones presented for the benchmark examples).

P.S. This issue is related to openjournals/joss-reviews#4902

Dear author,

The paper is written well, but some minor things can be improved.

• The JOSS guidelines require the summary to be “a clear description of the high-level functionality and purpose of the software for a diverse, non-specialist audience”. The current summary is not appropriate for non-specialist audiences and is a bit technical. You may consider moving the functionality and theory to another section and adding a more high-level introduction to GA and the developed library.
• “The software is therefore well suited for optimization tasks with computationally expensive fitness functions, where computation time is of the order of magnitude of seconds or more, like ab initio calculations in condensed matter physics." I wonder if there is any correlation between an expensive computation and the order of magnitude of the simulation time in computational sciences. I think this sentence needs a revision since any type of simulation (like a finite element simulation) can be expensive, making the fitness function an expensive expression to evaluate.
• For the shown example usages, it can be beneficial to mention explicitly that the figures can be reproduced by running example 1 and example 2. You also need to specify how example 2 should be run to produce the results presented in figure 2 (which may need a script on the repo to reproduce this figure).
• I think it is beneficial if you mention the limitations of the library briefly.

P.S. This issue is related to openjournals/joss-reviews#4902

Dear author,

The library has a set of small test cases with a Makefile intended to build and run these small test units. However, at the end of this process, it’s not clear whether the test is successful or not. As an improvement, I suggest that it reports to the user if each test passes or fails.

P.S. This issue is related to openjournals/joss-reviews#4902

There are many references in the JOSS paper the refer to the main branch of the code repository e.g. https://github.com/ttarkowski/quile/blob/main/tutorial/tutorial.txt and https://github.com/ttarkowski/quile/blob/main/README Since the main branch could be changed after the publication of the paper that could potentially create broken links. I think it would be better to create a tagged release of the code in github (the version of that release will be referred in the paper anyways) and create (permanent) links to files of that release.

In this section of the paper there is a mention in a search project in computational materials science (in ack the grant number is referred UMO-2016/23/B/ST3/03575) It would have been better to provide a link to that grant because it is not easy accessible through http://www.ncn.gov.pl/

Also regarding the sentence "PhD thesis (to be published on-line)", could you be more specific i.e. what is the title of the thesis, the author's name and the university. It is better to skip the note "to be published on-line" and instead create a bibliographic reference and cite that reference in the text.

Dear author,

The compiler requirement is elaborated in the tutorial file, but not enough information is provided in the readme file as the main documentation of the code. It took me some time to find out that I need GCC 10.2.1 to correctly build the examples. I first started with GCC 8 (with support for C++ 20 as described by the readme file), and then faced a couple of issues (like the #include <concepts> statement).

So, I suggest that you elaborate on the compiler-related things by bringing the relevant stuff from the tutorial file up to the readme file.

P.S. This issue is related to openjournals/joss-reviews#4902