Test to ensure transaction time-order fails for objects with Date index

[braverock]

Test to ensure transaction time-order fails for objects with Date index

[davidkleman]

Having some issues related to this. Are multiple transactions on the same Date supposed to work? In that case I'd like to repor the behaviour below. I tried converting manually to Posix but same behaviour. Apparently multiple transactions in one instrument goes through on one date (2007-08-08) but not the next day.

[1] "2007-07-20 02:00:00 FIRE -100 @ 23.4"
[1] "2007-07-23 02:00:00 WESC -50 @ 80"
[1] "2007-07-30 02:00:00 FIRE -25 @ 22.6"
[1] "2007-07-31 02:00:00 WESC -100 @ 80"
[1] "2007-08-08 02:00:00 CONS 100 @ 55.25"
[1] "2007-08-08 02:00:00 CONS 300 @ 55.25"
[1] "2007-08-08 02:00:00 SAND 200 @ 137.75"
[1] "2007-08-09 02:00:00 CONS -300 @ 54"
[1] "2007-08-09 02:00:00 CONS -300 @ 54"
Error in addTxn(myPortfolio.str, row$Ticker1st, TxnDate = row$PosixDate, :
Transactions must be added in order. TxnDate (2007-08-09 02:00:00) is before last transaction in portfolio (2007-08-09 02:00:00) for CONS

[braverock]

@davidkleman I'm not sure your issue is related to this (closed) bug.

In any event, if you have multiple transactions on the same Date, presumably you know the time the transaction occurred? Why not put in the transactions as it/they actually happened rather than throwing away the time information?

[davidkleman]

Yes, me neither. New to open source, am I supposed to write about this in some other way or place?

Unfortunately my broker only delivers the date of the transactions. As my use case is to analyze my manual trades which I haven't logged exact times for I don't know those.

In case someone encounters a similar situtation: Separating the transactions in time using row_number allows them to be added. E.g. stock_trans <- stock_trans %>% group_by(Ticker1st, PosixDate) %>% mutate(AdjDate = PosixDate + seconds(row_number())) %>% ungroup() %>% arrange(Ticker1st, AdjDate)

[braverock]

@davidkleman , see

?make.index.unique

for a much easier way to create a unique index without converting out of the xts time series and doing all the backflips you are doing.

In your example, assuming stock_trans is your time series of transactions, it would be simply

make.index.unique(stock_trans)

[davidkleman]

Neat, very neat - thank you. Backflips are dangerous.

[jaymon0703]

Safer in a swimming pool, but still dangerous

…

On Mon, 29 Oct 2018 at 11:15, David Kleman ***@***.***> wrote: Neat, very neat - thank you. Backflips are dangerous. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#51 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEI5vHjuPenhvHGIHDqKGizdpBBJjkVSks5upsc5gaJpZM4Nlc4H> .