Adding converted versions of all irregular datasets on Oscar-Oliveira/OR-Datasets

[JeroenGar]

I've converted all instances from OR-Datasets to the internal JSON format used here.

They are all in data/irregular/or-datasets. The ones that were already available (shirts, swim, trousers, shapes) are also included to be able to verify the conversion is correct.

In data/irregular_raw, there's just a link to the original files instead of copying them all.

[fontanf]

Hi,

Thank you for the contribution.

I prefer to get instances from the most primary source possible to reduce the probability of errors when successive conversions are applied. Here I think it is the ESICUP website https://www.euro-online.org/websites/esicup/data-sets/#1535972088237-bbcb74e3-b507

I also store the original instances in the repository (if not too big) because it often happens that the page hosting the files disappears; and in case there are errors in my conversion scripts, I can still easily check. For irregular packing, I store them in this directory https://github.com/fontanf/packingsolver/tree/master/data/irregular_raw

If these new instances are in the same format as the ones already in this repository, the conversion function is already implemented. One just needs to add a line for each new instance here

packingsolver/scripts/convert_irregular.py

Lines 693 to 698 in f0b82c6

	convert_oliveira2000(os.path.join("oliveira2000", "blaz_2007-04-23", "blaz.xml"))
	convert_oliveira2000(os.path.join("oliveira2000", "shapes_2007-04-23", "shapes0.xml"))
	convert_oliveira2000(os.path.join("oliveira2000", "shapes_2007-04-23", "shapes1.xml"))
	convert_oliveira2000(os.path.join("oliveira2000", "shirts_2007-05-15", "shirts.xml"))
	convert_oliveira2000(os.path.join("oliveira2000", "swim_2007-05-15", "swim.xml"))
	convert_oliveira2000(os.path.join("oliveira2000", "trousers_2007-05-15", "trousers.xml"))

And then running this scripts/convert_irregular.py script will generate the files in the data/irregular directory.

When instances come from a scientific article, I store them in a directory named as firstauthorYEAR to make it easier to retrieve where they come from. For example https://github.com/fontanf/packingsolver/tree/master/data/rectangle_raw By the way, that makes me think that it would be worth adding a doi.url containing the DOI of these articles in each of these directories.

[JeroenGar]

Oh, I was not aware you had a script in the repo do the conversion.
I completely agree with you about getting as close to the source as possible, but the XML files on the ESICUP website are not the originals.
And they do, as we discovered earlier, contain errors.
I have no idea who did the XML conversions and with what code.
The raw txt files on the website are the originals though.

There have actually been talks to move the entire ESICUP datasets page to GitHub because the site is not really being maintained at the moment.
It was only because my PhD supervisor is the current chair of ESICUP that I was able to quickly rectify the bug in the xml on the website.

We can discuss at ESICUP, I think it would be good to have your input on this.

[fontanf]

The .txt files are not available for all instances and don't contain the bin dimensions

I'm adding the missing instances.

Putting datasets on Github seems to be a good idea. It would make it easier to submit new dataset and to discuss