Hacking our way through the jungle

hippietrail

19:56

room topic changed to Hacking our way through the jungle: The intersection of Travel and Stack Overflow [travel-hacks]

@Roflcoptr come on in and paste your CSV so i can visualize it

Roflcoptr

Date,Views,Total Questions,Total Unanswered,Total Accepted,Total Answers,Total Comments,Total Votes,Total Badges,Total Users
8.11.2011 15:08,713.35,812,2,547,1793,3880,18981,2986,1221
8.11.2011 15:35,713.63,812,2,547,1793,3880,18984,2986,1222
8.11.2011 16:39,709.54,813,3,547,1793,3882,18990,2987,1222
8.11.2011 16:59,709.97,813,3,547,1793,3881,18998,2989,1223
8.11.2011 17:43,710.61,813,3,547,1793,3881,18998,2990,1224
8.11.2011 19:16,712.05,814,4,547,1793,3883,19020,2992,1225
8.11.2011 20:07,712.41,814,4,547,1793,3886,19025,2992,1225

hippietrail

ooh with headers thanks

Roflcoptr

yes

the newest version of the jar file should generate those headers automatically

i forgot to do that in the first version

hippietrail

seems better but needs tweaking

www-958.ibm.com/software/data/cognos/manyeyes/visualizations/…

Roflcoptr

yes

especially the y axis have to be adapted

hippietrail

20:08

gotta figure out the date format though

yeah i guess there's some way to set a range/scale for each doodad plotted on the same graph

i utterly lack the terminology for this

Roflcoptr

date format is european ;)

DD.MM.YYYY HH:MM

hippietrail

in your stuff or in manyeyes?

Roflcoptr

in my stuff

and in manyeyes

hippietrail

yeah i can grok it but manyeyes can't at least not without some nerding

hippietrail

20:30

i think many eyes doesn't have a way to do nonliner x axis

Roflcoptr

hmm that's not good

hippietrail

maybe there's a way to make it work with a scatter plot

aha yes scatterplot will work if we use a timestamp instead of a readable date

unless there's some way to get many eyes to convert the readable dates into ticks or something

Roflcoptr

hmm

i could also divide it by 100

the values

and the the scale has to by read by factor 100

i can't explain that in english correctly ;)

hippietrail

think you can add a column for timestamp / ticks?

many eyes will handle the scaling automatically so don't divide anything i reckon

Roflcoptr

Date,Views,Total Questions,Total Unanswered,Total Accepted,Total Answers,Total Comments,Total Votes,Total Badges,Total Users
8.11.2011 15:08,7.1335,812,2,547,1793,3880,18981,2986,1221
8.11.2011 15:35,7.1363,812,2,547,1793,3880,18984,2986,1222
8.11.2011 16:39,7.0954,813,3,547,1793,3882,18990,2987,1222
9.11.2011 16:47,7.1674,818,3,557,1809,3935,19234,3015,1235
9.11.2011 23:15,7.1917,820,3,559,1816,3940,19294,3024,1243
10.11.2011 10:32,7.2167,822,3,561,1818,3948,19381,3041,1247
10.11.2011 13:39,7.3082,823,3,561,1823,3957,19418,3054,1258

i thought like this

for the views

hippietrail

20:43

is something different?

ah i'll play with this but see if you can include date as just one number like a timestamp, ticks, seconds since system epoch, etc... whatever java likes best

Roflcoptr

yes

that's no problem

unix time

that is

milliseconds since 01.01.1970

hippietrail

yep ill see if it needs scaling or not

i can't figure out how to update the data or add a new version of it

Roflcoptr

^^

hmm

ill generate a new file

just for testing

Date,Views,Total Questions,Total Unanswered,Total Accepted,Total Answers,Total Comments,Total Votes,Total Badges,Total Users
1320958233068,736.27,827,5,564,1827,3983,19486,3072,1266
1320958245118,736.27,827,5,564,1827,3983,19486,3072,1266

now with unix time

hippietrail

scaling the views didn't make it any better for a normal line graph

www-958.ibm.com/software/data/cognos/manyeyes/visualizations/…

Roflcoptr

and if you scale it even more

between 0 and 1

hippietrail

20:56

no it doesn't like the new date field. maybe chop off seconds and smaller?

www-958.ibm.com/software/data/cognos/manyeyes/visualizations/…

Roflcoptr

thats not so easy

this are just milliseconds

passed since 01.01.1970

hippietrail

divide by (60 * 1000)?

Roflcoptr

yeah ok but then we can convert it back

not really

hippietrail

don't know how to import csv from the clipboard into excel either

Roflcoptr

u see the problem no

Date,Views,Total Questions,Total Unanswered,Total Accepted,Total Answers,Total Comments,Total Votes,Total Badges,Total Users
22015983,736.34,827,5,564,1828,3983,19486,3072,1266
22015983,736.34,827,5,564,1828,3983,19486,3072,1266

now it is somewhere in 1970s...

Sat, 12 Sep 1970 19:33:03 GMT

onlineconversion.com/unix_time.htm

hippietrail

21:05

have to save it as a file and then open it )-:

Roflcoptr

copy it in a text file

store it with file extension cvs

and then open it with excel

hippietrail

ick

Roflcoptr

does it work?

hippietrail

both rows are identical but let's see

Roflcoptr

yes

that's because i just run it twice

for testing purpose

hippietrail

21:13

ah of course we'd have to convert the original set into the new format. would require some spreadsheet expertise

but it still refuses to let me choose date )-: www-958.ibm.com/software/data/cognos/manyeyes/visualizations/…

Roflcoptr

strange

hippietrail

21:30

i'm becoming a grumpy hacker d-;

being too lazy to program it is proving to be hard work

Roflcoptr

what do you want to do?

hippietrail

get the nonlinear x axis to work

Roflcoptr

ah hmm

maybe we can use an algorithm

to choose a subset of the data set

that is more or less linear

hippietrail

i think a scatter plot is the way to go but the datetime is proving painfull

aha we could normalize it to start at 0

since it seems to be the magnitude it doesn't like

Roflcoptr

hmm

hippietrail

21:41

time to play with libreoffice

Roflcoptr

wahts that

hippietrail

bah libreoffice calc help talks about datebase but it's not in the app where the help says it should be

what they changed the name of openoffice to

well they forked it after the evil company became the owner of openoffice

Roflcoptr

ah i didn't know that

hippietrail

22:03

nup i just don't have a clue about spreadsheets

Roflcoptr

its not so diffuclt

just make a second column

hippietrail

there's a function to convert the dates from plain text to something but can't get it to work

Roflcoptr

and write a small formla

hippietrail

i just want to convert the column in place

sould be =DATEVALUE( ) but no clue what goes in ()

it's the minutiae i don't know

Roflcoptr

illl do it later

just have to figure out some weird grep stuff

hippietrail

22:08

no worries

Transcript for

Hacking our way through the jungle