Off topic:Need help on sorting on two columns

Question

[jon.brate]

I linearized a fasta file and counted the lengths of each sequence. Each line now consist of three tab separated columns, but column one also has a space in it.

>TCONS_00000098 gene=XLOC_000037    TGTGAACTGTTTGGAATGCCTAGATCATGATGAAGATTTTGGCGGCAAATCACGAACTACCAGATG    66
>TCONS_00000097 gene=XLOC_000037    TGTGAACTGTTTGGAATGCCTAGATCATGATGAAGATTTTGGCGGCAAATCACGAACTACCAGGTTGTGT    70
>TCONS_00000099 gene=XLOC_000037    TGAAGATTTTGGCGGCAA    18
>TCONS_00000100 gene=XLOC_000037    CAGATCGTCAAAAGTTTTTGAAGTTCCTCAAAAGAT    36
>TCONS_00000052 gene=XLOC_000022    AGCATTCG    8
>TCONS_00000025 gene=XLOC_000008    ACCGGTTTGCGTACTGATTTGCGTACTGGTTCGTGTA    37
>TCONS_00000132 gene=XLOC_000046    GTTTTAGTTGTTAGGTCTAACA    22
>TCONS_00000133 gene=XLOC_000046    CTGAGCAGTAACGCGACGCAGATCACTAAAGATCTG    36

I want to extract the longest isoform (TCONS...) of each gene, and I tried to sort the lines first on column 1, and then according to the lengths with the longest on top. I thought this command would work:

cat lengths.txt | sort -t '    ' -k1,1 -k3,3nr > sorted.txt

and it seems to somehow sort TCONS_00000097 right, but that is probably because of its name, not the length.

Output:

>TCONS_00000025 gene=XLOC_000008    ACCGGTTTGCGTACTGATTTGCGTACTGGTTCGTGTA    37
>TCONS_00000052 gene=XLOC_000022    AGCATTCG    8
>TCONS_00000097 gene=XLOC_000037    TGTGAACTGTTTGGAATGCCTAGATCATGATGAAGATTTTGGCGGCAAATCACGAACTACCAGGTTGTGT    70
>TCONS_00000098 gene=XLOC_000037    TGTGAACTGTTTGGAATGCCTAGATCATGATGAAGATTTTGGCGGCAAATCACGAACTACCAGATG    66
>TCONS_00000099 gene=XLOC_000037    TGAAGATTTTGGCGGCAA    18
>TCONS_00000100 gene=XLOC_000037    CAGATCGTCAAAAGTTTTTGAAGTTCCTCAAAAGAT    36
>TCONS_00000132 gene=XLOC_000046    GTTTTAGTTGTTAGGTCTAACA    22
>TCONS_00000133 gene=XLOC_000046    CTGAGCAGTAACGCGACGCAGATCACTAAAGATCTG    36