Skip to content

Millisecond timestamp from PgSql using DBI / RPostgreSQL ? Anyway to set up a default timeseries format for the timestamp type?

5 messages · Michael Lyszczek, Kent Hoxsey, Jeff Ryan

Original
Michael Lyszczek
# 
Kent,

Thanks for the response.

I am now doing this : 

select cast(extract(epoch from timestamp) as float) from trades order by timestamp;

this gives me a list that looks like .....
1353381185.089
 1353381185.089
  1353381189.96
  1353381189.96
  1353381189.96
 1353381190.353
 1353381190.353
 1353381190.353
 1353381191.139
 1353381191.139
 1353381191.139
  1353381256.04
 1353381855.805
  1353381864.52
 1353384140.324

Etc....

I try to bring it into R the same way, so new query:

NGZ2zoo <- dbGetQuery(con, "select cast(extract(epoch from timestamp) as float), instrument_id, trade_size, trade_price, exchange_generated_price, trade_id, agressor_side_isbuy from trades where trades.instrument_id = (select CAST(max(securities.\"UnderlyingSymbol\") as INT) from securities where securities.\"FNYShortName\" = 'iNGZ2')");
Summary(NGZ2zoo)
   date_part        instrument_id      trade_size     trade_price  
 Min.   :1.35e+09   Min.   :584445   Min.   : 2500   Min.   :3708  
 1st Qu.:1.35e+09   1st Qu.:584445   1st Qu.: 2500   1st Qu.:3742  
 Median :1.35e+09   Median :584445   Median : 2500   Median :3760  
 Mean   :1.35e+09   Mean   :584445   Mean   : 3070   Mean   :3759  
 3rd Qu.:1.35e+09   3rd Qu.:584445   3rd Qu.: 2500   3rd Qu.:3776  
 Max.   :1.35e+09   Max.   :584445   Max.   :70000   Max.   :3833  
 exchange_generated_price    trade_id        agressor_side_isbuy
 Mode :logical            Min.   :65001155   Min.   :-1.000     
 FALSE:2761               1st Qu.:65767044   1st Qu.:-1.000     
 TRUE :474                Median :66313346   Median : 0.000     
 NA's :0                  Mean   :66306294   Mean   :-0.068     
                          3rd Qu.:66880936   3rd Qu.: 1.000     
                          Max.   :67377417   Max.   : 1.000     

as.POSIXct(NGZ2zoo$date_part, origin="1970-01-01")
yields:
[3199] "2012-11-20 02:12:29 EST" "2012-11-20 00:29:30 EST"
[3201] "2012-11-20 02:15:27 EST" "2012-11-20 00:29:31 EST"
[3203] "2012-11-20 00:29:31 EST" "2012-11-20 00:29:33 EST"
[3205] "2012-11-20 02:15:28 EST" "2012-11-20 02:15:33 EST"
[3207] "2012-11-20 02:15:33 EST" "2012-11-20 02:16:11 EST"
[3209] "2012-11-20 02:18:21 EST" "2012-11-20 02:18:58 EST"
[3211] "2012-11-20 00:29:33 EST" "2012-11-20 00:29:34 EST"
[3213] "2012-11-20 00:29:40 EST" "2012-11-20 02:22:04 EST"
[3215] "2012-11-20 00:29:43 EST" "2012-11-20 02:22:31 EST"
[3217] "2012-11-20 02:22:33 EST" "2012-11-20 02:26:23 EST"
[3219] "2012-11-20 02:26:23 EST" "2012-11-20 02:27:53 EST"
[3221] "2012-11-20 02:29:30 EST" "2012-11-20 02:29:36 EST"
[3223] "2012-11-20 02:29:39 EST" "2012-11-20 02:47:05 EST"
[3225] "2012-11-20 02:47:05 EST" "2012-11-20 02:47:05 EST"
[3227] "2012-11-20 02:51:25 EST" "2012-11-20 02:53:37 EST"
[3229] "2012-11-20 02:53:44 EST" "2012-11-20 02:55:53 EST"
[3231] "2012-11-20 02:58:20 EST" "2012-11-20 03:14:16 EST"
[3233] "2012-11-20 03:24:15 EST" "2012-11-20 03:24:24 EST"
[3235] "2012-11-20 04:02:20 EST"


Even if I just do the system time:
[1] "2012-11-21 15:07:59 EST"


Is it just stored as a float? Or numeric? 

Any way you can see the actual value, not scientific notation?

Thanks so much,

Mike

-----Original Message-----
From: r-sig-finance-bounces at r-project.org [mailto:r-sig-finance-bounces at r-project.org] On Behalf Of Kent Hoxsey
Sent: Wednesday, November 21, 2012 3:03 PM
To: r-sig-finance at r-project.org
Subject: Re: [R-SIG-Finance] Millisecond timestamp from PgSql using DBI / RPostgreSQL ? Anyway to set up a default timeseries format for the timestamp type?

After working through a similar set of issues a while back, I discovered this very useful web page:

http://www.epochconverter.com/

Short summary: use your RDBMS to do the conversion to epoch time, and hand that number to R. Do not write SQL to write text to parse in R.

In my experience, the task of converting from a string representation to POSIXct representation in R can be quite, erm, tedious when you're working with a lot of data. In contrast, doing the conversion in a source database (Oracle for me) is really fast, as are the downstream operations in R.

Caveat: I have not attempted to preserve or work with the millis, so you might need to do some prototyping to get your conversion right coming out Postgres, but that web site includes quite a bit of discussion about the topic.

HTH.
On Nov 21, 2012, at 10:16 AM, Michael Lyszczek wrote:

            

      
      
      
    

        

      
      
    

        
_______________________________________________
R-SIG-Finance at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.

  
  


  


      

    

    

      
      

        


  

        

      Kent Hoxsey
    

    

      
      #
    

  


  

      
Ah, now it would appear you are asking simple R questions (rather than complicated finance questions) so I am going to point you toward some good sources. You will definitely want to do a little spelunking in the help, and there's a ton of good stuff (including examples) for working with time series on Stack Overflow (esp search for topics tagged with "r" and "xts").

Display of the number of digits is controlled by the options() function, read about it with help(options) and set it with something like: options(digits=10). You will want to read up on POSIXct(), particularly the "Sub-second Accuracy" section.

Your SQL  is most likely coming back into R as a dataframe (help(data.frame)) which you can look at using head() or tail() or str(). In combination with the digits settings, you will likely be able to see all of the digits you seek.

        
On Nov 21, 2012, at 12:22 PM, Michael Lyszczek wrote: