Avoid shadowing message in autoprint tests - #7446

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.07%. Comparing base (3c044ce) to head (1fabe00).
⚠️ Report is 6 commits behind head on master.

@@           Coverage Diff           @@
##           master    #7446   +/-   ##
=======================================
  Coverage   99.07%   99.07%           
=======================================
  Files          85       85           
  Lines       16609    16610    +1     
=======================================
+ Hits        16455    16456    +1     
  Misses        154      154           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

is the definition identical? if so we should start a deprecation cycle too

Ours might be faster when both arguments are character vectors; otherwise identical:

> `%notin%`
function (x, table)
match(x, table, nomatch = 0L) == 0L
<bytecode: 0x563bba77e7f8>
<environment: namespace:base>
> data.table::`%notin%`
function (x, table)
{
    if (is.character(x) && is.character(table)) {
        .Call(Cnotchin, x, table)
    }
    else {
        match(x, table, nomatch = 0L) == 0L
    }
}
<bytecode: 0x563bbab3aea0>
<environment: namespace:data.table>

I would definitely aim for deprecation, I don't think differences will be that big

Three CRAN packages import %notin%: bruceR, tidyfst, and weatherOz; the first two for re-exporting. Any clever tricks (à la cedta?) we might use so that interactive users who would get base::`%notin%` anyway will not get penalised? I see that some packages (e.g. backports) export objects conditionally on R version. Should we try to eventually use if (getRversion() < "4.6.0")) export("%notin%") and similarly emit deprecation warnings only on R ≥ 4.6.0?

An if() in the NAMESPACE might work... let's explore the full implications as a follow-up issue: #7453