I'm testing how useful language models are for supporting lesson maintenance. The fixes below were suggested by an LLM (Claude) after it reviewed this lesson's code: commands that no longer work or that behave differently from what the text describes, plus the text changes that follow from them. The commands were run against the Figshare dataset in bash 5.2 with GNU grep 3.11 and GNU coreutils 9.4 on Linux, and the counts match the lesson. The Windows points are based on how Git Bash behaves by default and weren't tested on Windows. Please treat these as suggestions. Feedback on whether this kind of review is useful is welcome.
Lesson: [Counting and mining research data with Unix](https://programminghistorian.org/en/lessons/research-data-with-unix). Line numbers refer to en/lessons/research-data-with-unix.md at 5bdece6.
- L51–57: the Figshare zip unpacks to
data/ with no proghist/ folder. Tell readers to create ~/proghist and unzip the file into it.
- L51–55, L67, L77, L93, L111:
c:\proghist\… fails in Git Bash because \ is an escape character. The replacement in the April 2025 note (c/Users/…) is missing its leading slash. Use ~/proghist/… on every platform and remove the note.
- L57, L67, L77, L93:
/user/USERNAME/… and ~/users/USERNAME/… don't exist. Use ~/proghist/….
- L87–89:
wc -c counts bytes and wc -m counts characters on every platform, so the "OS X/Linux use -m" note is inaccurate. Use -m in the text and note that -c counts bytes. The two give the same number for this ASCII data.
- L107, L111, L113: when grep searches several files, it puts the filename at the start of each line, which corrupts the first column of the saved subsets. Add
-h (grep -ih …, grep -ivh …), explain it in one line, and add it to the Summary (L122).
- L111–113: saving the output to a
.csv file doesn't convert it, and the contents are still tab-separated. Reword the text: spreadsheet programs use the file extension to decide how to split columns.
I'm testing how useful language models are for supporting lesson maintenance. The fixes below were suggested by an LLM (Claude) after it reviewed this lesson's code: commands that no longer work or that behave differently from what the text describes, plus the text changes that follow from them. The commands were run against the Figshare dataset in bash 5.2 with GNU grep 3.11 and GNU coreutils 9.4 on Linux, and the counts match the lesson. The Windows points are based on how Git Bash behaves by default and weren't tested on Windows. Please treat these as suggestions. Feedback on whether this kind of review is useful is welcome.
Lesson: [Counting and mining research data with Unix](https://programminghistorian.org/en/lessons/research-data-with-unix). Line numbers refer to
en/lessons/research-data-with-unix.mdat5bdece6.data/with noproghist/folder. Tell readers to create~/proghistand unzip the file into it.c:\proghist\…fails in Git Bash because\is an escape character. The replacement in the April 2025 note (c/Users/…) is missing its leading slash. Use~/proghist/…on every platform and remove the note./user/USERNAME/…and~/users/USERNAME/…don't exist. Use~/proghist/….wc -ccounts bytes andwc -mcounts characters on every platform, so the "OS X/Linux use-m" note is inaccurate. Use-min the text and note that-ccounts bytes. The two give the same number for this ASCII data.-h(grep -ih …,grep -ivh …), explain it in one line, and add it to the Summary (L122)..csvfile doesn't convert it, and the contents are still tab-separated. Reword the text: spreadsheet programs use the file extension to decide how to split columns.