Question

对于R来说，我是新手。我正在查看从以下内容返回的RODBC结果：

> library(RODBC)
> dbcon <- odbcDriverConnect("DRIVER={SQL SERVER};SERVER=MYSERV;DATABASE=SOME", tabQuote='', colQuote='')
> sqlQuery(dbcon, "SELECT 3, 'a', ' ', '', NULL")
    .1 .2 .3 .4
1 3  a NA NA NA

为什么' '和''返回NA？我在文档中找不到这种行为的任何内容。我错过了什么？

Answer 1

因此我决定查看RODBC sqlQuery方法的源代码，因为我记得它是开源的：

sqlQuery <-
    function(channel, query, errors = TRUE, ..., rows_at_time)
{
    if(!odbcValidChannel(channel))
       stop("first argument is not an open RODBC channel")
    if(missing(query))
        stop("missing argument 'query'")
    ## could argue that 'max' should restrict rows_at_time
    rows_at_time <- if(missing(rows_at_time)) attr(channel, "rows_at_time")
    else max(1, min(1024, rows_at_time))
    stat <- odbcQuery(channel, query, rows_at_time)
    if(stat == -1L) {
        if(errors) return(odbcGetErrMsg(channel))
        else return(invisible(stat))
    } else return(sqlGetResults(channel, errors = errors, ...))
}

所以它会调用sqlGetResults：

sqlGetResults <-
    function (channel, as.is = FALSE,
              errors = FALSE, max = 0, buffsize = 1000,
              nullstring = NA_character_, na.strings = "NA",
              believeNRows = TRUE, dec = getOption("dec"),
              stringsAsFactors = default.stringsAsFactors())
{
    if(!odbcValidChannel(channel))
       stop("first argument is not an open RODBC channel")
    as.df <- function(value, colnames) {
        for(i in seq_along(value))
            if(is.list(value[[i]])) class(value[[i]]) <- "ODBC_binary"
        ## convert list to data frame
        class(value) <- "data.frame"
        names(value) <- make.unique(colnames)
        row.names(value) <- seq(along=value[[1L]])
        value
    }
    cols <- .Call(C_RODBCNumCols, attr(channel, "handle_ptr"))
    ## FIXME: should this be <= 0L?
    if(cols < 0L) {
        if(errors) return("No data")
        else return(invisible(-1L))
    }
    cData <- .Call(C_RODBCColData, attr(channel, "handle_ptr"))
    dbdata <- odbcFetchRows(channel,
                            max = max,
                            buffsize = buffsize,
                            nullstring = nullstring,
                            believeNRows = believeNRows)
    if(dbdata$stat < 0L) {
    if(errors) return(odbcGetErrMsg(channel))
    else return(invisible(dbdata$stat))
    }
    data <- as.df(dbdata$data, cData$names)
    if(nrow(data) > 0L) {
        cols <- ncol(data)
        enc <- attr(channel, "encoding")
        if(length(na.strings))
            for (i in 1L:cols)
                if(is.character(data[,i]))
                    data[data[,i] %in% na.strings, i] <- NA
        if(is.logical(as.is)) {
            as.is <- rep(as.is, length = cols)
        } else if(is.numeric(as.is)) {
            if(any(as.is < 1 | as.is > cols))
                stop("invalid numeric 'as.is' expression")
            i <- rep(FALSE, cols)
            i[as.is] <- TRUE
            as.is <- i
        } else if(length(as.is) != cols)
            stop("'as.is' has the wrong length ", length(as.is),
                 " != cols = ", cols)
        for (i in seq_len(cols)) {
            if(is.character(data[[i]]) && nchar(enc))
                data[[i]] <- iconv(data[[i]], from = enc)
            if(as.is[i] || is.list(data[[i]])) next
            if(is.numeric(data[[i]])) next
            if(cData$type[i] == "date")
                data[[i]] <- as.Date(data[[i]])
            else if(cData$type[i] == "timestamp")
                data[[i]] <- as.POSIXct(data[[i]])
            else
                data[[i]] <- type.convert(as.character(data[[i]]),
                                          na.strings = na.strings,
                                          as.is = !stringsAsFactors,
                                          dec = dec)
        }
    }
    data
}

因此对于字符串/ varchar值，所有这些都归结为：

data[[i]] <- type.convert(as.character(data[[i]]),
                                         na.strings = na.strings,
                                          as.is = !stringsAsFactors,
                                          dec = dec)

所以让我们试试type.convert：

> type.convert("a")
[1] a

好的，这是预料之中的。让我们现在尝试奇怪的案例：

> type.convert("")
[1] NA

嗯......好的，这就是我们得到NA

的方式

> type.convert("       ")
[1] NA

好吧，那也是NA。现在，as.is为什么不返回NA？

if(as.is[i] || is.list(data[[i]])) next

啊哈，它只是在设置as.is时退出，并且不会调用type.convert() 这解释了为什么在设置NA标志时它不会返回as.is

RODBC：为什么sqlQuery（）中的空值和空格值的值“NA”？

1 个答案: