Command-line techniques used on this project, with the why. Newest topics added
as they come up.
Checking whether a product/package name is taken
When naming Sorepoint we needed to know if candidate names were free. Two cheap
checks from the terminal, no browser needed for the first one.
1. Is the npm package name free? — npm view
npm view sorepoint namenpm view <pkg> <field> (aliases: npm info, npm show) asks the npm registry
for a package’s metadata without installing anything. Ask for one field
(name, version, description) to keep output tiny.
- If the package exists, it prints that field.
- If it doesn’t exist, npm exits with an error containing the phrase
is not in this registry(a 404).
So we can turn that into a FREE/TAKEN flag by grepping for the error text:
npm view sorepoint name 2>&1 | grep -q 'is not in' && echo FREE || echo TAKENPiece by piece:
2>&1— redirect stderr into stdout so the pipe sees the error text too
(npm prints errors to stderr; without this,grepwould never see them).grep -q 'is not in'—-q= quiet: print nothing, just exit0if the
pattern is found (name is free) or1if not (name exists).A && B || C— shell short-circuit: runBifAsucceeded, elseC. Here:
found the “not in registry” error →echo FREE, otherwise →echo TAKEN.
(Careful with this idiom in general: ifBitself fails,Cstill runs. Fine
here becauseechonever fails.)
2. Check a whole list at once — a for loop
for n in sorepoint triagely firstfix onepitch weakspot; do
printf "%-12s " "$n"
npm view "$n" name 2>&1 | grep -q 'is not in' && echo FREE || echo TAKEN
donefor n in a b c; do … done— loop over a literal list,$nis each item.printf "%-12s " "$n"— print the name left-justified in a 12-char column
(-= left-justify,12= width,s= string) so the FREE/TAKEN column lines
up in a neat table.printfis better thanechohere because it doesn’t add a
newline, letting the loop body finish the line.
3. npm-free is necessary but NOT sufficient
A free npm name only means no npm package. A name can still be owned by a real
company (trademark, .com, existing SaaS). So we paired every npm check with a web
search — that’s how kerbside (npm-free) was ruled out: there’s already a
“Kerbside” booking-software product. Registry check = cheap first filter; web
search = the real clearance.
Scaffolding into a non-empty repo — moving files aside with mv
create-next-app refuses to run in a directory that already contains
non-standard files: it only tolerates a small allow-list (.git, .gitignore,
LICENSE, a few IDE/config files). Our repo already had file.md and docs/,
so the generator would have bailed with “the directory contains files that
could conflict.”
Rather than scaffold into a throwaway subfolder and merge hidden files back by
hand, the cleaner move is: park the conflicting files, scaffold in place, put
them back.
mv file.md docs "$SCRATCH"/mv a b c dest/—mvtakes many sources and one final destination. When
the last argument is a directory, every preceding argument is moved into it.
So this moves bothfile.mdand thedocs/folder into$SCRATCH.- Order matters: the last path is always the destination.
mv a brenames
a→b;mv a b dir/moves both intodir/. The trailing/is a habit that
makes intent obvious and makesmverror loudly ifdirisn’t actually a
directory (safer than silently renaming).
Why plain mv, not git mv
git mv looks like a file operation but is really a git operation — it
stages the rename in the index. For temporary plumbing like this (move out, move
back, net-zero change) we deliberately use plain mv so git’s index is never
touched. Afterward git status just shows the files as working-tree deletions:
git status --short
# D docs/PLAN.md
# D docs/SPEC.md
# D file.mdThe leading space + D means deleted in the working tree but not staged —
git’s first column is the staged (index) status, the second is the
working-tree status. Since we move the files right back before anything is
committed, this reverses to a clean tree with no history churn.
Software engineering
Infrastructure & security
Computer science
Hardware & engineering