问题
We keep all our laboratory data in a Sybase database. When I want to do data manipulation and analysis I read the data into R with RODBC.
library(RODBC)
channellab <- odbcConnect("Labdata")
indivs <-sqlQuery(channellab,'SELECT * from CGS.Specimen')
So far so good, except that CGS.Specimen is a table for our entire lab holdings. There are almost 40000 specimens (rows) with 66 variables. It takes an unnecessary amount of time to read especially when I am only interested in the holdings of one study which has about 1000 specimens. The obvious way to fix it would be
indivs <-sqlQuery(channellab,'SELECT * from CGS.Specimen WHERE StudyCode="RP"')
But when I run that I get
[1] "42S22 -143 [Sybase][ODBC Driver][Adaptive Server Anywhere]Column not found: Column 'RP' not found" "[RODBC] ERROR: Could not SQLExecDirect 'SELECT * from CGS.Specimen WHERE StudyCode=\"RP\"'"
I have played with quotation marks and I have played with escape slashes but alas I have gotten nowhere.
for instance
indivs <-sqlQuery(channellab,'SELECT * from CGS.Specimen WHERE StudyCode= 'RP' ')
Error: unexpected symbol in "indivs <-sqlQuery(channellab,'SELECT * from CGS.Specimen WHERE StudyCode= 'RP"
indivs <-sqlQuery(channellab,"SELECT * from CGS.Specimen WHERE StudyCode= 'RP' ")
Error in .Call(C_RODBCFetchRows, attr(channel, "handle_ptr"), max, buffsize, : negative length vectors are not allowed
What do you suggest?
回答1:
The last error you report is from the .Call function to the c function RODBCFetchRows. Thus your problem with the quotation marks appears to be solved. However, with the said error you still have a problem. Perhaps the rows are not correctly reported (the c function tries to allocate vectors based on this value, but does not check for negative values). Try:
indivs <-sqlQuery(channellab,"SELECT * from CGS.Specimen WHERE StudyCode= 'RP'", believeNRows = FALSE)
回答2:
Edited:
indivs <-sqlQuery(channellab,'SELECT * from CGS.Specimen WHERE StudyCode=\'RP\'')
Ok can you run another query where some data is returned?
WHERE StudyCode=StudyCode for example? Does this give the same error about -ve lengths?
this means that the above is correct for escaping the StudyCode. The -Ve lenght is another error. Can you return not all columns with * but specific columns where there are no nulls?
回答3:
Try option believeNRows=FALSE. This worked for me on a similar error (the negative length error).
This question was more to the point: Querying Oracle DB from Revolution R using RODBC
来源：https://stackoverflow.com/questions/3622042/why-will-where-statement-not-work-in-sqlquery-in-rodbc