一段时间后Rterm.exe崩溃

Question

I've written a R-script using the tidyquant (and therefore dplyr Version: 0.7.5) way of data handling. My script loads about 1000 ( num_etry ) entries from a HANA DB, aggregates them into a bunch of ts , fills missing values, computes the seasonal component of the data and writes it back into the HABA DB. Something similar to https://cran.rstudio.com/web/packages/sweep/vignettes/SW01_Forecasting_Time_Series_Groups.html . Since the code works only with the given Database and is from work (Copyright and data security and stuff) I can't produce a MWE (sorry).

Running this script in Rstudio leads to an "R Session Abborted" error.
Running it within the Rgui leads to the RGui just vanishing from the screen.
Running it within the R.exe produces at least an Windows error log entry:

    <EventData>
      <Data>Rterm.exe</Data> 
      <Data>3.43.8260.0</Data> 
      <Data>5a27b6f6</Data> 
      <Data>R.dll</Data>
      <Data>3.43.8260.0</Data> 
      <Data>5a27b6ea</Data> 
      <Data>c0000094</Data> 
      <Data>00000000000c1ea6</Data> 
      <Data>2228</Data> 
      <Data>01d42e352023186a</Data> 
      <Data>C:\Users\Public\R-3.4.3\bin\x64\Rterm.exe</Data> 
      <Data>C:\Users\Public\R-3.4.3\bin\x64\R.dll</Data> 
      <Data>be0ec2de-9b19-11e8-aad7-d4bed926ba7c</Data> 
    </EventData>

At first I thought of it as a "too little memory" kind of issue (have a look at Using R with tidyquant and massiv data ) but no matter how big or small I make the num_etry var the interpreter itself crashes. With smaller values it just takes longer (leaking memory within dplyr?). The hard part is that the script sometimes takes several hours before crashing. Of course I can restart the script manually, but this is a rather dull procedure and there is no way I can run the script over the weekend. My R-script is located on an network drive, maybe this is the cause?

At the moment I am at a loss at how to make my script work.

Do you have any idea, hint, anything to help me out?

Answer 1

For Compatibility reason I used R in the 3.4.3 version. Now I'm using the newer 3.5.1 version, which works quite fine. So apparently there was some kind of internal issue that got fixed now.