stack dsl

[tj]

just realized it doesn't really make sense to use commits at the command level in some cases, for example:

mkdir /tmp/redis
curl -L# http://download.redis.io/releases/redis-stable.tar.gz | tar -zx --strip 1 -C /tmp/redis
cd /tmp/redis && make install

If you were to change the release tarball it would re-execute but make install would not, so what we really need is:

redis {
  tmp {
    curl -L# http://download.redis.io/releases/redis-stable.tar.gz | tar -zx --strip 1
    make install
  }
}

where "redis {}" is the stack and "tmp {}" is just an example directive

[wilmoore]

The alternative way to solve this (if you wanted to avoid the DSL) is to put the redis related lines in a new file, then call/source it from the main script. You'd commit the SHA of the external script contents.

Not saying this is necessarily better; but could potentially be less to maintain.

On the other hand, writing DSLs is always fun, so there's that :)

[tj]

haha yeah that's just an example, lots of use-cases where a few commands should be considered a single commit

[wilmoore]

😄 - Right on.

[l3pp4rd]

you could probably treat group of commands as related, without two newlines, example:

# one command, hashed with newlines as block
mkdir /tmp/dir
cd /tmp/dir
echo /hello/world

# command two, could also reset some state, like **pwd**
pwd # /home/user

The group could be executed in sh -c "group of commands" in order to restore EXPORTs and other state..