some thoughts about patch in packaging

sed

1. easy to get, so old that stable, one of GNU toolkit

2. is turing complete, so in theory has high limitation

1. very hard to use, learn and maintain. simple usage like sed -i 's/a/b/g' is already hard to read. If with regex, it’s hell

patch

1. easy to get

2. easy to generate, just diff a b > patch

3. easy to use, patch -Np1 < patch

1. stick to the origin file, not possible to apply to many files.

2. if many replaces, patch will be huge, even if they are the same rules.

ast-grep

1. easy to use

2. easy to generate

1. not built-in

2. only support some languages

3. added by author: pre-compiled version doesn’t support glibc < 2.31, or you can compile it yourself.

Example

code base need to change a function name

1. sed, just replace, but will replace ALL, even in string. The problem of using part of text is that

2. patch, patch may be huge, but is accepted

2. ast-grep, replace the function call

Now consider the update of software update

• sed: need to change even if var names have changed.

• patch: every time repatch.

• ast-grep: just keep the same replace, if the api and code shape doesn’t change. Even if changed, still easy to write.