Re: [Gperf-bugs] Gperf 2.7.2 test file questions

[Bruno Haible]

On Friday 22 February 2002 23:41, Bruce Lilly wrote:

> Anybody know why the test files in the gperf 2.7.2
> distribution use suboptimal keys?

Probably because the people who determined the key positions had to do
this by trial-and-error ;-)

> In particular,
> tests/c-parse.gperf uses "1,3,$"; "2,6,$" would
> be better and "1,2,6" would be better still.

How do you measure which is better? gperf-2.8 will use "2,4,5" in this
case by default, and "2,4,5" is just as good as "1,2,6".

> The tests/cplusplus.gperf example uses "1,4,7,$"; key
> position 7 is redundant as "1,4,$" suffices, and
> "1,3,5" is optimal.

Indeed. gperf-2.8 will use the key positions "1,3,5" in this case.

Bruno