Fortran 90读取含冒号的CSV文件报错，如何解决？

报错信息：

forrtl: severe (59): list-directed I/O syntax error, unit 21, file /data3/khee/data/AWS/ASCII/STATION/2021/csvfile/test.csv
Image              PC                Routine            Line        Source             
a.out              000000000043B9EE  Unknown               Unknown  Unknown
a.out              0000000000412AB0  Unknown               Unknown  Unknown
a.out              00000000004036D3  Unknown               Unknown  Unknown
a.out              00000000004034DE  Unknown               Unknown  Unknown
libc-2.17.so       000000333901F84D  __libc_start_main     Unknown  Unknown
a.out              00000000004033E9  Unknown               Unknown  Unknown

原代码：

program read_data

  implicit none

  integer :: HSID, DSID
  real :: HLAT, HLON, HALT, DTEMP, DWDEG, DWSPD, DPRCP, DSPRS, DMSLP, DHMD, DSR, DSRH
  character(len=16) :: DATE

  open(21,file='/data3/khee/data/AWS/ASCII/STATION/2021/csvfile/test.csv',form='formatted',access='sequential',status='old')

10  read(21,*,end=90) DSID, DATE, DTEMP, DWDEG, DWSPD, DPRCP, DSPRS, DMSLP, DHMD, DSR, DSRH
           write(*,*) DSID, DATE, DTEMP, DWDEG, DWSPD, DPRCP, DSPRS, DMSLP, DHMD, DSR, DSRH
           
go to 10
90 continue

  write(*,*) "EOF"
  stop
  
end program read_data

输入行示例：

848,2021-01-01 00:00,-4.1,9.1,2,0,,,,0,0

问题本质 #

你使用的列表式I/O（read(21,*)）默认将空格、逗号、制表符等都视为字段分隔符。DATE字段2021-01-01 00:00中的空格会被当成分隔符，导致程序仅将2021-01-01读入DATE变量，后续的00:00会被当作下一个字段，但该字段需要匹配实数类型的DTEMP，冒号无法被解析为实数，最终触发语法错误。冒号是表象，核心是列表式读取的分隔规则与CSV结构不匹配。

解决方案 #

方案一：修改OPEN语句，指定仅用逗号作为分隔符

这是改动最小的方法，只需在OPEN语句中添加delim='comma'选项，强制列表式I/O仅将逗号视为字段分隔符，空格和冒号会被保留在字符字段中：

open(21,file='/data3/khee/data/AWS/ASCII/STATION/2021/csvfile/test.csv',&
     form='formatted',access='sequential',status='old', delim='comma')

修改后，原有的read(21,*,end=90)语句可以正常工作：DATE字段会完整读取2021-01-01 00:00，后续字段也能正确匹配对应类型。

方案二：先读取整行，再用内部文件解析

如果CSV格式更复杂（比如字段包含引号或嵌套逗号），可以先将整行读取到字符变量，再通过内部文件指定格式解析，精准控制每个字段的读取规则：

program read_data
  implicit none

  integer :: DSID
  real :: DTEMP, DWDEG, DWSPD, DPRCP, DSPRS, DMSLP, DHMD, DSR, DSRH
  character(len=16) :: DATE
  character(len=256) :: line  ! 存储整行内容，长度可根据实际调整

  open(21,file='/data3/khee/data/AWS/ASCII/STATION/2021/csvfile/test.csv',&
       form='formatted',access='sequential',status='old')

10 read(21, '(A)', end=90) line  ! 读取整行到字符变量
  ! 用内部文件解析，格式字符串匹配CSV的逗号分隔结构
  read(line, '(I0, ",", A16, ",", F0.0, ",", F0.0, ",", I0, ",", F0.0, ",", &
              & F0.0, ",", F0.0, ",", F0.0, ",", I0, ",", I0)') &
              & DSID, DATE, DTEMP, DWDEG, DWSPD, DPRCP, DSPRS, DMSLP, DHMD, DSR, DSRH
  write(*,*) DSID, DATE, DTEMP, DWDEG, DWSPD, DPRCP, DSPRS, DMSLP, DHMD, DSR, DSRH
           
go to 10
90 continue

  write(*,*) "EOF"
  stop
  
end program read_data

说明：I0和F0.0表示自动匹配整数/实数的长度，","用于跳过字段间的逗号，空字段会被解析为对应类型的默认值（如实数0、整数0）。

内容的提问来源于stack exchange，提问作者Khee