Zsh Mailing List Archive
Messages sorted by: Reverse Date, Date, Thread, Author

Named reference (and other) pending work



Pending patches about named references (from most to least important)

Remove support for named references to subscripted variables - workers/54363
Fix nameref issues in builtin "unset" - workers/54364
  - Implement references to positional parameters - workers/54365
Detect invalid variable names in reference initializers - workers/54278
Remove unnecessary tag based reference loop detection - workers/54277

Other pending patches (from most to least important)

Run ZERR trap also when there was a Zsh error - workers/54374
Remove partial duplication of strsetfn and arrsetfn - workers/54340
  - Refine string and array slice assignments - workers/54348
Only ever turn global parameters into named directories - workers/54328

Review of dereferencing lookups

I have started reviewing all parameter lookups that also dereference named references, i.e., more or less all calls to realparamtab->getnode(). Several of these lookups have issues. In many cases, they should simply not perform any dereferencing (i.e., call getnode2() instead of getnode) but a few are more problematic. I have already reviewed most lookup and will start sending out patches and/or start discussions to figure out what to do.

Review of typeset implementation

A few months ago, I started an exhaustive review of the implementation of typeset and found many issues, which lead to many side tracks. Some of them have been resolved and a few others are among the patches listed above. However, I know for sure that there are still issues with typeset. In particular with -p and -m when applied to named references. I suspect that there are also still issues with some calls to typeset p=v where p is an existing named reference.

Turn placeholder dereferencing into an error

In ksh, any attempt to dereference a placeholder reference triggers an error. Zsh should do the same. Currently it sometimes exhibits rather dubious behaviors in place of triggering an error.

Discuss the semantics of references to not-yet-defined variables

With the current implementation, references to not-yet-defined variables lead sometimes to behaviors that are nonsensical in my opinion. I think that a different approach could lead to references that are not only less problematic but also generally more useful.

Other issues

I'm aware of a number of other issues but as long as named references are released as an experimental feature, I think that they can be addressed after the release.

Philippe



Messages sorted by: Reverse Date, Date, Thread, Author