Annotations are a bit difficult. The problem is the cleanup process. The cleanup process is being executed before any annotation script is run. This leads to problems if the cleanup process actually removes files from hard disk, which actually need to be around for the annotation process e.g. filesize checks for generated files. There are two options to solve the problem.

Problem

Let the annotation be the following

[run.dd.annotations]
filesize = "ls -al {output} | awk '{{print $5}}'"
modified = "ls -al /tmp/Cargo.toml.dd | awk '{{print $6\" \"$7\" \"$8}}'"

which returns the file size and the last modification date of the file. The filesize can not be computed afterwards, since {output} refers to a parameter from hyperfine.

Solutions

• Execute the cleanup process externally
  This is not possible if the annotation command refers to parameters delivered by hyperfine
• Attach annotation process with the cleanup (i.e. &&)
  Integration to cleanup can be done attaching the commands. But a cli tool for manipulation of json files is necessary e.g. for filesize

   jq --arg filesize $(ls -al /tmp/Cargo.toml.dd | awk '{{print $5}}') '.results[0].filesize=$filesize' out.json  > cat out.json