This topic contains 4 replies, has 4 voices, and was last updated by Max Kozlov 7 months, 1 week ago.

AuthorPosts

March 13, 2017 at 11:28 pm #66321
As a newbie to manipulating CSV files within PowerShell I need a foot up with the following problem.
stats.csv contains data for a time series of two objects A,B. I need to create a report referencing some values from this and also need to calculate new values from the data it contains.
stats.csv available here:
The first value on each row is the time series (PT). There are two rows for each time event in the file, one for each object (ON) A and B. The second value for each row is the object (ON). As you can see if you scroll through the CSV, there will be two entries for each PT – one for ON A and another for ON B. Each ON has it's own unique data for values for the PT for TT,RIO,TB,RS,WS,RT that need to be calculated or passed directly to a new CSV.
The following calculations are needed for the report (excuse the crudeness below):
$total_tt = sum TT where PT value is the same for ON=A And ON=B
e.g. for the same PT entries @ 02/06/2017 15:45
$total_tt would be 12804.49537$total_tb = sum TB where PT value is the same for ON=A And ON=B
e.g. for the same PT entries @ 02/06/2017 15:45
$total_tb would be 180.226005$avg_rp = (if $total_tt>0(sum RIO where same PT for ON=A and=B)/$total_tt)
e.g. for PT entries @ 02/06/2017 15:45
$avg_rp would be 53.81826021 (percentage)$avg_rs = (if sum of RT from same PT value for ON=A And ON=B>0(RIO*RS for ON=A+RIO*RS where ON=B)/sum of RIO for A and B)
e.g. for PT entries @ 02/06/2017 15:45
$avg_rs would be 9.944376086$avg_ws = within the same TS(if $total_ttRIOARIOB>0((TTARIOA)*WSA+(TTBRIOB)*WSB/(RIOARIOB)
e.g. for PT entries @ 02/06/2017 15:45
$avg_ws would be 17.33865279$avg_rt = average RT where same PT value is observed for ON=A And ON=B
e.g. for PT entries @ 02/06/2017 15:45
$avg_rt would be 0.293554Create a 1st output file: report_detail.csv with the following values (some newly calculated, some from the original csv):
PT,$total_tt,$total_tb,$avg_rp,$avg_rs,$avg_ws,$avg_rt,RT(ON:A from stats.csv),RT(ON:B from stats.csv)
Next, return the row numbers from report_detail.csv that contain the maximum and 95th percentile value for $total_tt and send them to report_headlines.csv with first row headers for the csv:
measurement,PT,$total_tt,$total_tb,$avg_rp,$avg_rs,$avg_ws,$avg_rt,RT(ON:A from original CSV),RT(ON:B from original CSV).
Any help is appreciated to get me started.

March 13, 2017 at 11:59 pm #66330
You could start with "reading" your csv file with ImportCSV. The you iterate through all lines with a Foreach loop and calculate whatever you need. In the end you can output all this to the console or export it to a new csv with ExportCSV
BTW: A few lines of your csv file would have been enough to show what it's about! ðŸ˜‰

March 14, 2017 at 12:17 am #66331
Hi. No issues. I can trim back the csv ðŸ˜‰
Would you be able to show me a working example of what you suggest of what I'm asking for with $total_tt? If I get a foot up I can work the remainder.


March 14, 2017 at 1:28 am #66340
Here you go:
# region Input $Data = ImportCsv .\stats.csv #endegion #region Process $A = ($Data  group ON  ? { $_.Name eq 'A' }).Group $B = ($Data  group ON  ? { $_.Name eq 'B' }).Group $Analysis = foreach ($ONA in $A) { RemoveVariable ONB ErrorAction SilentlyContinue $ONB = $B  ? { $_.PT eq $ONA.PT } if ($ONB) { $SumTT = [decimal]$ONA.TT + [decimal]$ONB.TT $SumRT = [decimal]$ONA.RT + [decimal]$ONB.RT NewObject TypeName PSObject Property @{ PT = [DateTime]$ONA.PT RTA = [decimal]$ONA.RT RTB = [decimal]$ONB.RT SumTT = $SumTT SumTB = [decimal]$ONA.TB + [decimal]$ONB.TB AvgRB = $( if ($SumTT gt 0) { ([decimal]$ONA.RIO + [decimal]$ONB.RIO)/$SumTT } else { 0 } ) 'AvgRB(%)' = $( if ($SumTT gt 0) { '{0:P2}' f (([decimal]$ONA.RIO + [decimal]$ONB.RIO)/$SumTT) } else { 0 } ) AvgRS = $( if ($SumRT gt 0) { (([decimal]$ONA.RIO * [decimal]$ONA.RS) + ([decimal]$ONB.RIO * [decimal]$ONB.RS)) / ([decimal]$ONA.RIO * [decimal]$ONB.RIO) * 100 } else { 0 } ) AvgWS = $( if ($SumTT  ([decimal]$ONA.RIO + [decimal]$ONB.RIO)) { (([decimal]$ONA.TT[decimal]$ONA.RIO)*[decimal]$ONA.WS + ([decimal]$ONB.TT[decimal]$ONB.RIO)*[decimal]$ONB.WS) / ([decimal]$ONA.RIO[decimal]$ONB.RIO) } else { 0 } ) AvgRT = ([decimal]$ONA.RT+[decimal]$ONB.RT)/2 } } } $Analysis = $Analysis  Sort SumTT Descending #endregion #region Output $Analysis  select PT, SumTT, SumTB, AvgRB, AvgRS, AvgWS, AvgRT, RTA, RTB  ExportCsv .\report_detail.csv NoType Force $Index95 = [math]::ceiling($Analysis.Count * 0.95) $95PercentileValue = ($Analysis  Sort SumTT)[$Index95] $Maximum = $Analysis  select First 1 $Maximum  AddMember MemberType NoteProperty Name 'Measurement' Value 'Maximum' $Maximum  select Measurement, PT, SumTT, SumTB, AvgRB, AvgRS, AvgWS, AvgRT, RTA, RTB  ExportCsv .\report_headlines.csv NoType Force $95PercentileValue  AddMember MemberType NoteProperty Name 'Measurement' Value '95Percentile' $95PercentileValue  select Measurement, PT, SumTT, SumTB, AvgRB, AvgRS, AvgWS, AvgRT, RTA, RTB  ExportCsv .\report_headlines.csv NoType Append #endregion
You owe me $50..lol

March 14, 2017 at 6:51 am #66345
I think than grouping will be much more effective if
instead of grouping by ON we directly group by PT$Data = ImportCsv D:\stats.csv  GroupObject Property PT  ForeachObject { $a = $_.Group  WhereObject { $_.ON eq 'A' } $b = $_.Group  WhereObject { $_.ON eq 'B' } [PSCustomObject]@{ PT = $_.Name A = $a B = $b } #manipulate data 1 } #manipulate data 2

AuthorPosts
You must be logged in to reply to this topic.