R 示例：日志分词器

LogTokenizer 转换函数会从表中读取可变长字符串，即日志消息。然后，它会标记每条日志消息的字符串，以返回每个标记。

您可以在 Vertica Github 存储库中找到更多 UDx 示例：https://github.com/vertica/UDx-Examples。

加载函数和库

创建库和函数。

=> CREATE OR REPLACE LIBRARY rLib AS 'log_tokenizer.R' LANGUAGE 'R';
CREATE LIBRARY
=> CREATE OR REPLACE TRANSFORM FUNCTION LogTokenizer AS LANGUAGE 'R' NAME 'LogTokenizerFactory' LIBRARY rLib FENCED;
CREATE FUNCTION

使用函数查询数据

以下查询显示了如何使用 UDTF 运行查询。

=> SELECT machine,
          LogTokenizer(error_log USING PARAMETERS spliton = ' ') OVER(PARTITION BY machine)
     FROM error_logs;
 machine |  Token
---------+---------
 node001 | ERROR
 node001 | 345
 node001 | -
 node001 | Broken
 node001 | pipe
 node001 | WARN
 node001 | -
 node001 | Nearly
 node001 | filled
 node001 | disk
 node002 | ERROR
 node002 | 111
 node002 | -
 node002 | Flooded
 node002 | roads
 node003 | ERROR
 node003 | 222
 node003 | -
 node003 | Plain
 node003 | old
 node003 | broken
(21 rows)

UDTF R 代码


LogTokenizer <- function(input.data.frame, parameters.data.frame) {
  # Take the spliton parameter passed by the user and assign it to a variable
  # in the function so we can use that as our tokenizer.
  if ( is.null(parameters.data.frame[['spliton']]) ) {
    stop("NULL value for spliton! Token cannot be NULL.")
  } else {
    split.on <- as.character(parameters.data.frame[['spliton']])
  }
  # Tokenize the string.
  tokens <- vector(length=0)
  for ( string in input.data.frame[, 1] ) {
    tokenized.string <- strsplit(string, split.on)
    for ( token in tokenized.string ) {
      tokens <- append(tokens, token)
    }
  }
  final.output <- data.frame(tokens)
  return(final.output)
}

LogTokenizerFactory <- function() {
  list(name    = LogTokenizer,
       udxtype = c("transform"),
       intype  = c("varchar"),
       outtype = c("varchar"),
       outtypecallback=LogTokenizerReturn,
       parametertypecallback=LogTokenizerParameters)
}


LogTokenizerParameters <- function() {
  parameters <- list(datatype = c("varchar"),
                     length   = c("NA"),
                     scale    = c("NA"),
                     name     = c("spliton"))
  return(parameters)
}

LogTokenizerReturn <- function(arg.data.frame, parm.data.frame) {
  output.return.type <- data.frame(datatype = rep(NA,1),
                                   length   = rep(NA,1),
                                   scale    = rep(NA,1),
                                   name     = rep(NA,1))
  output.return.type$datatype <- c("varchar")
  output.return.type$name <- c("Token")
  return(output.return.type)
}