notes.billmill.org / computer_usage / http_commands /

curl

last updated: Oct 20, 2023

I use curl very frequently, I'm kind of shocked I didn't have a page for it yet.

Download files in parallel

I have previously done this with wget, xargs, or both. Today I decided to figure out how to use curl's native parallel download support.

The impetus for the test was wanting to download all the images from the planet.com
gallery to use as background images, and the full script is here.

The gist is that you create a series of commands for curl to run (curl has no support for files containing newline-separated URLs) that looks like this:

url = https://example.com/uri/1

url = https://example.com/uri/2

and so on and so forth. You can specify additional options; it's documented here, but for this task I had no need to do so.

Once I generated this file, I downloaded it with:

curl --remote-name-all --output-dir images \
    --parallel --parallel-immediate --parallel-max 5 \
    --continue-at - \
    --write-out "%{response_code} -> %{filename_effective}\n" \
    --config "$URLS"

The parallel options are documented here; I used --continue-at - to prevent curl from re-downloading files; it lacks the ability to do so by filename so this does it by requests to the server instead. (It stinks that curl lacks an equivalent of wget's --no-clobber)

The --write-out bit tells curl to output something for each file downloaded; read the docs for the options of what you can write.

curl converter is handy for turning curl commands into various other formats

websocket in curl is a new-ish capability. Currently only usable via the API, but command line usage is planned.

caching

TIL that you can use --etag-compare and --etag-save to download files with curl only if the file has changed. So,

curl --etag-compare etag.txt --etag-save etag.txt 'https://csvbase.com/ngkabra/Statistics' -O

says:

compare the etag on the server to etag.txt
save the etag from the server if the file is downloaded as etag.txt
download https://csvbase.com/ngkabra/Statistics
-O means "save the file with a filename similar to the one on the remote", in this case Statistics

The result is that curl will download the file only if it hasn't already been downloaded, or if the remote etag changes.

Via Why my favourite API is a zipfile on the European Central Bank's website, more specifically the hn comments (but the post is worth reading)

or -z

curl also has the -z option, which accepts a filename. If you give it a date it can parse, or a filename, it will only download the remote file if newer than -z. So this command means "only download the zip file if it's newer than the /tmp/euro.zip file we have on disk":

curl -s -o /tmp/euro.zip -z /tmp/euro.zip https://www.ecb.europa.eu/stats/eurofxref/eurofxref-hist.zip

Backlinks:

↑ up