What I threw together below is I think very much like the response Bepa gave, although I didn't see it until I refreshed the page before submitting this. The code below seems to work, and includes sample data (including data to cause the multiple-city temperature case, which your original data, where Dallas was at 23, would not). I know it can probably be done better, however, and I would hope other, more experienced monks would comment so we all might learn from their experience. Good luck in finding your solution.

begin 664 test8.pl
M(R$O=7-R+V)I;B]P97)L("UW"G5S92!S=')I8W0["G5S92!W87)N:6YG<SL*
M"FUY("5R97-U;'1S(#T@*"`B;6EN(B`]/B`M,2P@(FUA>"(@/3X@,2`I.PIM
M>2`H)6UI;FUA>"P@)&QI;F4L("1L;V-A;&4L("1C:71Y+"`D<W1A=&4L("1R
M97-U;'0L($!T96UP<RD["@IW:&EL92`H)&QI;F4@/2`\1$%403XI('L*"6-H
M;VUP*"1L:6YE*3L*"2@D;&]C86QE+"!`=&5M<',I(#T@<W!L:70H+UQS*R\L
M("1L:6YE*3L*"2@D<W1A=&4L("1C:71Y*2`]('-P;&ET*"\L+RP@)&QO8V%L
M92P@,BD["@H)>PH)"2,@4V]R="!T:&4@=&5M<&5R871U<F5S"@D);7D@0'-T
M96UP<R`]('-O<G0@>R1A(#P]/B`D8GT@0'1E;7!S.PH)"4!T96UP<R`]($!S
M=&5M<',["@E]"@H)9F]R96%C:"!M>2`D=&5M<"`H<7<H;6EN(&UA>"DI('L*
M"0DD<F5S=6QT(#T@,3L*"0EI9B`H)'1E;7`@97$@(FUA>"(I('L*"@D)"2,@
M4F5V97)S92!T:&4@=&5M<&5R871U<F4@87)R87D@:68@;&]O:VEN9R!F;W(@
M"@D)"2,@("!T:&4@;6%X('1E;7!E<F%T=7)E+"!S;R!W92!O;FQY(&AA=F4@
M=&\@;&]O:R`*"0D)(R`@(&%T(&5L96UE;G0@,"X*"0D)0'1E;7!S(#T@<F5V
M97)S92A`=&5M<',I.PH)"7T*"@D)(R!/;FQY(&-O;7!A<F4@86=A:6YS="!T
M:&4@8W5R<F5N="!V86QU92!I9B!T:&5R92!I<R!O;F4*"0EI9B`H97AI<W1S
M*"1M:6YM87A[)'-T871E?7LD=&5M<'U[(FQO8V%T:6]N(GTI*2!["@H)"0DC
M(%)E='5R;G,@+3$L(#`L(#$@9&5P96YD:6YG(&]N(&AO=R!T:&4@=&5M<"!C
M;VUP87)E<R`*"0D)(R`@('=I=&@@=&AE(')E8V]R9&5D(&]N90H)"0DD<F5S
M=6QT(#T@)'1E;7!S6S!=(#P]/B`D;6EN;6%X>R1S=&%T97U[)'1E;7!]>R=D
M96=R965S)WT["@H)"0DC($5N<W5R97,@=&AA="!A(&YO;BUN96=A=&EV92!V
M86QU92!W:6QL('1R:6=G97(*"0D)(R`@(&-H86YG:6YG('9A;'5E<PH)"0DD
M<F5S=6QT("H]("1R97-U;'1S>R1T96UP?3L*"0E]"@H)"6EF("@D<F5S=6QT
M(#X](#`I('L*"@D)"2,@0F5C875S92!T:&4@87)R87D@:7,@;W)D97)E9"`H
M86YD(')E=F5R<V5D+"`*"0D)(R`@(&9O<B!M87@I+"!T:&4@=F%L=64@=V4@
M<V5E:R!I<R!T:&4@9FER<W0@96QE;65N=`H)"0DD;6EN;6%X>R1S=&%T97U[
M)'1E;7!]>R=D96=R965S)WT@/2!S:&EF="A`=&5M<',I.PH*"0D)(R!)9B!T
M:&4@=F%L=64@=V%S(&YO="!T:&4@<V%M92!A<R!T:&4@<F5C;W)D960@;VYE
M+`H)"0DC("`@<F5P;&%C92!I=#L@;W1H97)W:7-E+"!A9&0@:70@=&\@=&AE
M(&QI<W0@;V8*"0D)(R`@(&-I=&EE<PH)"0EI9B`H)')E<W5L="D@>PH)"0D)
M0'LD;6EN;6%X>R1S=&%T97U[)'1E;7!]>R=L;V-A=&EO;B=]?2`]("@@)&-I
M='D@*3L*"0D)?2!E;'-E('L@"@D)"0EP=7-H*$![)&UI;FUA>'LD<W1A=&5]
M>R1T96UP?7LG;&]C871I;VXG?7TL("1C:71Y*3L*"0D)?0H)"7T*"7T)"GT)
M"@HC(%!R:6YT('1H92!S=&%T92P@9F]L;&]W960@8GD@=&AE(&UA>"!A;F0@
M;6EN('1E;7!E<F%T=7)E<RP@96%C:"!F;VQL;W=E9`HC("`@8GD@=&AE(&-I
M='D@;W(@8VET:65S(&EN('=H:6-H(&ET(&]C8W5R<F5D+"!W:71H(&UU;'1I
M<&QE(&-I=&EE<PHC("`@<V5P97)A=&5D(&)Y("(L("(*9F]R96%C:"!M>2`D
M<W1A=&4@*&ME>7,H)6UI;FUA>"DI('L*"7!R:6YT*"1S=&%T92P@(B`B*3L*
M"69O<F5A8V@@;7D@)'1E;7`@*'%W*&UA>"!M:6XI*2!["@D)<')I;G0H)'1E
M;7`L("(@=&5M<"`B*3L*"0EP<FEN="@D;6EN;6%X>R1S=&%T97U[)'1E;7!]
M>R=D96=R965S)WTL("(@(BD["@D)<')I;G0H:F]I;B@B+"`B+"!S;W)T*$![
M)&UI;FUA>'LD<W1A=&5]>R1T96UP?7LG;&]C871I;VXG?7TI*2P@(B`B*3L*
M"7T*"7!R:6YT*")<;B(I.PI]"E]?1$%405]?"D-A+$9R965M;VYT(#@W(#0U
M(#DQ(#4V(#@Y(#4T(#8W(#0U(#<V(#0S(#@Q(#0X(#<V(#0S(`I#82Q"97)K
M;&5Y(#4V(#,T(#8W(#,X(#8V(#,T(#4V(#,Q(#8W(#0U(#<Q(#0U(#8U(#0R
M(`I4>"Q(;W5S=&]N(#(X(#,P(#,Q(#,Q(#,T(#,V(#4Q(#<R(#<U(#<U(#<W
M(#<Y(#@Q(#@X"E1X+$1A;&QA<R`R."`R."`S,B`T,R`T-2`T-2`T-B`W-B`W
1."`W.2`X,R`X-"`X-B`X-PH`
`
end
[download]

Update: After finding shortly after posting this that there was a high probability this was indeed homework (compare with Sort hash of hashes), I was quite angry, and in my anger posted it to both, thinking perhaps they might at least get caught in their attempt to dupe those willing to help. Perhaps it was poor etiquite of me to post the solution on both questions, and if so, I appologize to the other monks herein. For a while, I considered removing my code altogether, but was unsure if that would be appropriate either. I am grateful for the sage advice of tilly, jeffa, and others, who suggested another course of action. The block that has now replaced the code above is actually the code, as I tested it, UUEncoded. Perhaps it only slows down those who would take advantage of the willingness to help found here, perhaps not. It is my hope that damage to trust by those involved is realized. It is my hope that damage to trust by those involved is realized. It is sad to think that now, with each post I read and consider, the question in my mind will be, "is this a real request for help, or homework?"