The project folder as received

The folder 20261023_esrf_datuak_irene held three things: the archive edited-files-hc6528_400_4.zip (34.1 MB), the folder it had been extracted to, edited-files-hc6528_400_4 2/, and a Finder .DS_Store. Josu's brief: study the extracted folder in depth, leave the zip alone, keep the files as they are, rename the folder to originals, and make a properly named copy of every file in a new folder.

ls -la
ls "edited-files-hc6528_400_4 2" | wc -l
du -sh "edited-files-hc6528_400_4 2"

The files could not be read: mode 000 from the extraction

The first attempt to read a file was refused with Permission denied. The folder was drwx---rwx with a com.apple.quarantine attribute, and every file inside was ---------- (mode 000): the archive had been extracted without read permission for anyone, owner included. The owner is the local user, so the mode could be changed. This is the only change made to the originals: names, contents, sizes and modification times are untouched (P13 proves it with checksums).

whoami; id
ls -la@ | head
chmod 644 *.xye

What a file contains: three columns, no header

The head and tail of the first scan (…900C…-0s.xye) were printed. Every line is three numbers: 2θ in degrees, an intensity and an estimated standard deviation (esd). There is no header, no comment line and no metadata of any kind inside the files. The first lines, below about 0.02°, have zero intensity with a large esd: the beamstop's shadow. The 2θ axis runs from 0.008° to 26.999°.

head -25 hc6528_400_4_Qu_900C_w0003_b0001-0s.xye
tail -3  hc6528_400_4_Qu_900C_w0003_b0001-0s.xye

    0.008000  0.0000000      1567.0943
    0.009000  0.0000000      100.84378
    …
    0.023000  1690.1763      412.41486
    …
   26.999000  82.555925      3.5181567
So the "contents" from which the copies were to be named are the numbers alone; all the metadata is in the
filenames and in the file system (P06).

Two sizes, two line counts

The sizes and line counts of all 101 files fall into exactly two groups.

ls -l *.xye | awk '{print $5}' | sort | uniq -c
wc -l *.xye | awk '{print $1}' | sort | uniq -c

The filename pattern decoded

Every name matches hc6528_400_4_Qu_<tag>_w0003_b0001-<N>s.xye. Read as: hc6528 the ESRF proposal number (HC-6528), 400_4 the sample, Qu a sample label whose meaning is still to be asked, <tag> a nominal temperature written either as 900C … 1300C (with a C) or as 1225 … 1300 (bare), or background_nosample; w0003 a parameter of the data reduction not identified from the files alone; b0001 a bin width of 0.001°, which the data's step confirms; <N>s seconds elapsed since the first scan. The tags are not unique: 39 files share the tag 1270, 14 the tag 1300; only the elapsed seconds tell them apart.

Per-file statistics, sorted by elapsed time

A Python pass over every file computed: number of points, first and last 2θ, lines that did not parse, the integrated intensity, the maximum intensity, and the file's modification time. The table was sorted by the elapsed seconds in the name, which put the experiment in order: a 25 °C-step ramp from 900 °C, then dense series at 1240–1300 °C, then the background.

python3 - <<'EOF'
# for every *.xye: parse the three columns, read the elapsed seconds from the name, read the mtime,
# print sec, tag, n, x0, x1, bad lines, sum(I), max(I), mtime; sort by sec
EOF

The modification times are the scan times

The modification time of every file minus its elapsed seconds gives the same instant, 2026-06-10 18:09:48, for all 101 files. So the clock time of every scan is recoverable, and the two sources of time (name and file system) agree exactly. The experiment ran from 18:09 on 10 June 2026 to 04:46 on 11 June, 10.6 hours.

Grid, counting statistics and a crude peak count

A second pass checked the 2θ step of every file (min and max of the differences), computed the median relative error e/y above 2° as a proxy for counting time, and counted peaks: local maxima more than 6 esd above a 301-point rolling median, at 2θ above 1°. The peak count is a consistency check between scans, not a peak search.

python3 analyze.py    # the site's own tool: it repeats this pass and every checksum on each build

The naming scheme for the copies

Since the files carry no metadata inside, the new names were built from what was recovered: the order, the temperature, the elapsed time, and the one property the contents themselves give, the exposure class. Decided: hc6528_400_4_Qu_<seq>_T<temp>C_t<elapsed>s_<class>.xye, with seq 001–101 zero-padded so that alphabetical order is acquisition order, the temperature zero-padded to four digits (the originals mixed 900C and 1270), the elapsed seconds zero-padded to five digits, and class either long or short. The background is hc6528_400_4_Qu_101_background_nosample_t38192s_long.xye. The C suffix of the originals is not in the name but is recorded in the index; what it meant is still to be asked.

The folder renamed to originals

mv "edited-files-hc6528_400_4 2" originals

The new folder renamed/, one copy per original

A Python script read the mapping (P09), copied each file with shutil.copy2 into renamed/ under its new name, and computed the SHA-256 of source and copy before moving to the next file, stopping on the first mismatch. None occurred.

mkdir renamed
python3 - <<'EOF'
# for each original: new name from (seq, temp, sec, class); shutil.copy2; assert sha256(src) == sha256(dst)
# write file_index.csv with both names, metadata, statistics and the hash
EOF

The index and the README

file_index.csv at the project root maps every original to its copy, with the nominal temperature, the original tag and whether it had the C suffix, the elapsed seconds, the clock start time, the point count, the 2θ range, the exposure class, the median relative error, the maximum intensity above 1° and the SHA-256. README.md describes the layout, the format, the naming scheme, the experiment's stages, what the data show and the caveats.

Independent re-verification of all 101 pairs

A separate script re-read file_index.csv, recomputed the SHA-256 of every original and every copy, and compared both with the hash in the index.

python3 - <<'EOF'
# for each row: sha256(originals/orig) == sha256(renamed/new) == row.sha256
EOF

This site, and the record of every step

Josu asked for the steps to be logged for the group, and for the project's website in the family's form. The site is ecosystem/: data files, a generator and a Quarto project with the family design kit, as thermomat-phd-offers and rietveld-minimisation. Plotting the contents was the next phase and was not started then, as asked: "do not do, for the moment".