关于 UNIX Grep 命令

Question

我需要编写一个 shell 脚本来选择 /exp/files 目录中的所有文件（不是目录）。对于目录中的每个文件，我想查找是否收到文件的最后一行。文件中的最后一行是预告片。最后一行的第三个字段是数据记录数，即 2315（文件中的总行数 -2 (header,trailer) ）。在我的 unix shell 脚本中，我想通过检查 T 来检查最后一行是否是预告片记录，并想检查文件中的行数是否等于 (2315+2)。如果这成功了，那么我想将文件移动到不同的目录 /exp/ready。

tail -1 test.csv 
T,Test.csv,2315,80045.96

同样在输入文件中，有时 0 或 1 个预告记录字段可以在双引号内

"T","Test.csv","2315","80045.96"
"T", Test.csv, 2212,"80045.96"
T,Test.csv,2315,80045.96

Answer 1

您可以使用以下内容测试最后一行是否存在：

tail -1 ${filename} | egrep '^T,|^"T",' >/dev/null 2>&1
rc=$?

此时，如果行以T, 或"T", 开头，则$rc 将为0，假设这足以捕获预告片记录。

确定后，您可以使用以下方法提取行数：

lc=$(cat ${filename} | wc -l)

您可以通过以下方式获得预期行数：

elc=$(tail -1 ${filename} | awk -F, '{sub(/^"/,"",$3);print 2+$3}')

比较两者。

因此，将所有这些结合在一起，这将是一个好的开始。它会输出文件本身（我的测试文件num[1-9].tst）以及指示文件是否正常或为什么不正常的消息。

#!/bin/bash
cd /exp/files
for fspec in *.tst ; do
    if [[ -f ${fspec} ]] ; then
        cat ${fspec} | sed 's/^/   /'
        tail -1 ${fspec} | egrep '^T,|^"T",' >/dev/null 2>&1
        rc=$?
        if [[ ${rc} -eq 0 ]] ; then
            lc=$(cat ${fspec} | wc -l)
            elc=$(tail -1 ${fspec} | awk -F, '{sub(/^"/,"",$3);print 2+$3}')
            if [[ ${lc} -eq ${elc} ]] ; then
                echo '***' File ${fspec} is done and dusted.
            else
                echo '***' File ${fspec} line count mismatch: ${lc}/${elc}.
            fi
        else
            echo '***' File ${fspec} has no valid trailer.
        fi
    else
        ls -ald ${fspec} | sed 's/^/   /'
        echo '***' File ${fspec} is not a regular file.
    fi
done

示例运行，显示我使用的测试文件：

   H,Test.csv,other rubbish goes here
   this file does not have a trailer
*** File num1.tst has no valid trailer.
   H,Test.csv,other rubbish goes here
   this file does have a trailer with all quotes and correct count
   "T","Test.csv","1","80045.96"
*** File num2.tst is done and dusted.
   H,Test.csv,other rubbish goes here
   this file does have a trailer with all quotes but bad count
   "T","Test.csv","9","80045.96"
*** File num3.tst line count mismatch: 3/11.
   H,Test.csv,other rubbish goes here
   this file does have a trailer with all quotes except T, and correct count
   T,"Test.csv","1","80045.96"
*** File num4.tst is done and dusted.
   H,Test.csv,other rubbish goes here
   this file does have a trailer with no quotes on T or count and correct count
   T,"Test.csv",1,"80045.96"
*** File num5.tst is done and dusted.
   H,Test.csv,other rubbish goes here
   this file does have a traier with quotes on T only, and correct count
   "T",Test.csv,1,80045.96
*** File num6.tst is done and dusted.
   drwxr-xr-x+ 2 pax None 0 Feb 23 09:55 num7.tst
*** File num7.tst is not a regular file.
   H,Test.csv,other rubbish goes here
   this file does have a trailer with all quotes except the bad count
   "T","Test.csv",8,"80045.96"
*** File num8.tst line count mismatch: 3/10.
   H,Test.csv,other rubbish goes here
   this file does have a trailer with no quotes and a bad count
   T,Test.csv,7,80045.96
*** File num9.tst line count mismatch: 3/9.

Answer 2

如果你想在文件被写入并关闭后移动它们，那么你应该考虑使用类似 inotify、incron、FAM、gamin 等的东西。

Answer 3

这段代码通过一次调用 awk 来完成所有的逻辑计算，这使得它非常高效。它还 NOT 硬编码 2315 的示例值，而是使用预告片行中包含的值，因为我相信这是您的意图。

如果您对结果满意，请记得删除echo。

#!/bin/bash

for file in /exp/files/*; do
  if [[ -f "$file" ]]; then
    if nawk -F, '{v0=$0;v1=$1;v3=$3}END{gsub(/"/,"",v0);exit !(v1 == "T" && NR == v3+2)}' "$file"; then
      echo mv "$file" /ext/ready
    fi
  fi
done

更新

我必须添加{v0=$0;v1=$1;v3=$3}，因为 SunOS 的 awk 实现不支持 END{} 可以访问字段变量（$0、$1、$2 等），但如果您必须将其保存到用户定义的变量中想在 END{} 内处理它们。查看This awk feature comparison link中第一个表的最后一行

Answer 4

这里没有 UNIX shell，但是

#!/bin/bash
files=$(find /exp/files -type f)

应该把所有文件放在一个 BASH 数组中；然后按照上面建议的 paxdiablo 遍历它们中的每一个应该让你排序

Answer 5

destination=/exp/ready
for file in /exp/files/*.csv
do
    var=$(tail -1 "$file" | awk -F"," '{ gsub(/\042|\047/,"") }
    $1=="T" && $3 == "2315" { print "ok" }')
    if [ "$var" = "ok" ]; then
        echo mv "$file" "$destination"
    else
        echo "invalid: $file"
    fi
done

Answer 6

#!/bin/bash

ex findready.sh <<'HERE'
  i#!/bin/bash/

  let NUMLINES=$(wc -l $1)
  let TRAILER=$(cat $1 | tail -1 | tr -d '"' | sed 's/^\(.\).*$/\1/')

  if [[ $NUMLINES -eq 2317 && $TRAILER == "T" ]] ; then
      mv $1 /exp/ready/$1
  fi
  .
  wq
HERE

chmod a+x findready.sh

find /exp/files/ -type f -name '*.csv' -exec ./findready.sh {} ';' > /dev/null 2>&1