Desktop

[brett_walton]

I am looking for logic to determine how to pull the following data 

I have a fact table that has transaction date and time and computer per transaction the second table is the device table showing who was logged into the device (there is no log out). My intended logic will be to link date, and device ID and then find the time that is the first time found less the log in time from the fact table . So the first entry I want to determine BOB was logged and second ROB

                     Date      Time        Device

Example :      1/1/17   08:10:53   RFGN3550            

                     1/1/17   10:13:40   RFGN3560

I have a device Log table that says who was logged into that device 

                 Log in Date    Time       Device          Person

Example:           1/1/17   08:00:00   RFGN3550     BOB 

                         1/1/17  09:00:00    RFGN3550     MARY  

                         1/1/17  09:00:00    RFGN3560     SAM

                         1/1/17  10:00:00    RFGN3550    ROB

[OwenAuger · Accepted solution]

This seems to have been deemed an M question, but here is one of those "dime a dozen" DAX solutions:

PBIX sample

 Calculated column in Transactions:

Person = 
VAR TransactionDateTime = Transactions[DateTime]
RETURN
    CALCULATE (
        //Arbitrary tie-breaker
        FIRSTNONBLANK ( Logins[Person], 0 ),
        CALCULATETABLE (
            LASTNONBLANK ( Logins[DateTime], 0 ),
            Logins[DateTime] <= TransactionDateTime
        )
    )

Owen

[Sean · Accepted solution]

@brett_walton

I suspect you've modified @OwenAuger's formula to only look at your Time columns!

Note that he checks a Date/Time column.

So to address this create a Date-Time Column in both tables like this

Date-Time = [Date]&" "&[Time]

Both of these columns will default to Text - so you need to convert them to Data Type - Date/Time - again in both tables.

This should resolve it!

Good Luck!

[MarcelBeug]

Maybe DAX would be more appropriate, but in M (Edit Queries) it can be done with the following code.

I combined dates and times in each table to the field "DateTime".

Steps:

• Merge table Logins with Transactions (Logins are nested tables in Transctions table)
• Add a column with nested Logins with DateTime <= Transaction.DateTime
• Remove original Logins column
• Select latest Logins
• Expand the nested table, returning only the name
• Apply type text to the Name 

let
    Source = Table.NestedJoin(Transactions,{"Device"},Logins,{"Device"},"Logins",JoinKind.LeftOuter),
    AddedLoginsBeforeTransaction = Table.AddColumn(Source, "Logins before Transaction", (CT) => Table.SelectRows(CT[Logins], each [DateTime]<=CT[DateTime])),
    RemovedLogins = Table.RemoveColumns(AddedLoginsBeforeTransaction,{"Logins"}),
    LatestLogin = Table.TransformColumns(RemovedLogins, {{"Logins before Transaction", each Table.SelectRows(_, let latest = List.Max(_[DateTime]) in each _[DateTime] = latest)}}),
    NameFromLogin = Table.ExpandTableColumn(LatestLogin, "Logins before Transaction", {"Person"}, {"Person"}),
    Typed = Table.TransformColumnTypes(NameFromLogin,{{"Person", type text}})
in
   Typed

 By the way, the second is SAM, not ROB.

[Sean]

Folks it seems @MarcelBeug maybe giving our undisputed M expert @ImkeF a run for her money!

[MarcelBeug]

Thanks @Sean, @ImkeF is aware.

[Sean]

@MarcelBeugCan you post a video again? Thanks!

Like you did here...

http://community.powerbi.com/t5/Desktop/Reorder-data-with-Power-Query-repeating-words/m-p/118860#M50...

Thanks!

[MarcelBeug]

@Sean Especially with Power Query I think that videos have enormous added value, as code is typically generated by clicking the right buttons on the menus.

A good example in https://community.powerbi.com/t5/Desktop/Parsing-Embedded-JSON/m-p/118765#M50133 in which the solution is basically simple and straightforward, but it requires a lot of clicks.

Even with a video it takes 3 minutes to explain.

[ImkeF]

Happy to hand-over :-)

[Sean]

@ImkeFif we can poll this community one thing will be certain!

When it comes to DAX there are many so called "experts"

But when it comes to M everyone will vote for you!

So don't hand over anything yet!

[MarcelBeug]

When it comes to theoretical M knowledge, I think @ImkeF can more or less compete by now.

With regard to practical experience I'm definitely lagging behind.

But I have that screencasting software as a card in my sleeve.

And my userid is with a capital M...

[OwenAuger · Accepted solution]

This seems to have been deemed an M question, but here is one of those "dime a dozen" DAX solutions:

PBIX sample

 Calculated column in Transactions:

Person = 
VAR TransactionDateTime = Transactions[DateTime]
RETURN
    CALCULATE (
        //Arbitrary tie-breaker
        FIRSTNONBLANK ( Logins[Person], 0 ),
        CALCULATETABLE (
            LASTNONBLANK ( Logins[DateTime], 0 ),
            Logins[DateTime] <= TransactionDateTime
        )
    )

Owen