On Sat, May 12, 2018 at 12:04 AM <ati...@mail.ccsf.edu> wrote:
> How do you recommend I check errors. I dismiss them because I don't like
> if err !=nil { log.error(err.error()}
> being all over my project. Suggestions please, Thanks!
>
>
Code have a positive path on which the normal operation happens. But code
also have a negative path when something fails. There might be a gap
between these two paths, and robust code closes this gap to zero.
If you don't handle the negative path, then the gap widens to be
"everything not on the positive path". This makes the program brittle as it
might do wrong things where it should have reported an error.
Errors are values, and you should think about passing them up the call
stack, eventually wrapping them as they traverse up to the caller from the
callee. Local logging tells the log something is wrong, but the caller
might know better and have a good workaround. Don't assume you know how the
caller is going to treat an error.
Other languages have tried handling the negative path by means of
exceptions, but the risk of doing so is that the programmer might either
forget to handle the error, or be sloppy while doing so.
In my experience, errors are a gritty detail of many function calls, and
they should be treated with the same respect as your positive "happy" path.
Software is only correct when:
- Valid inputs are accepted with a valid result, and
- Invalid inputs are rejected with an error
There is a direct argument to Type I/II errors in statistics: you want to
control both types of errors. Not passing errors as values severely hampers
your ability to eventually detect Type II errors.
- Valid inputs are accepted with an invalid result, and (Type I)
- Invalid inputs are not rejected and produce undefined results (Type II)
--
You received this message because you are subscribed to the Google Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
