min (resp. max) calculation wrong if station has no negative (resp. positive) temperature

[setop]

go-1brc/1brc.awk

Line 6 in fafba32

mins[$1] = $2 < mins[$1] ? $2 : mins[$1]

In this statement, if station has never been encountered,mins[$1] is initialized to 0. If station has no negative value, min stays at zero, producing wrong result. Same for maxs.

$ wget https://github.com/benhoyt/go-1brc/raw/master/1brc.awk

$ cat << EOF > m.txt 
A;1
A;4
A;2
EOF

$ gawk -F\; -f 1brc.awk m.txt 
{A=0.0/2.3/4.0}  # wrong, min is 1

Any valid solution must test if a station already has an entry or not.

BTW, nice post. I enjoyed reading about optimizing go code using profiling.

[benhoyt]

Yeah, good point. Would you like to propose a PR that does if (counts[$1]) ... or similar?

[shadowrun96]

Great! why don't you show us your own fix and the benchmark results?
But yeah, I had a passing thought that you need to check against a range and not a single value, but I was too tired by the end of it, and moved on to finish the rest of the AWK book instead :-)