> On Tue, Apr 30, 2019 at 5:44 AM Tom Lane <tgl@sss.pgh.pa.us> wrote: >> FWIW, my recommendation for this sort of thing is almost always >> to not use CHAR(n). The use-case for that datatype pretty much >> disappeared with the last IBM Model 029 card punch. ...
Perhaps the "tip" on the character datatype page (https://www.postgresql.org/docs/11/datatype-character.html) should be updated as the statement "There is no performance difference among these three types..." could easily lead a reader down the wrong path. The statement may be true if one assumes the planner is able to make an optimal choice but clearly there are cases that prevent that. If the situation is better explained elsewhere in the documentation then just a link to that explanation may be all that is needed.