Fortran跨目录读写数据问题：可导入但无法完成导出验证

Hey Rob, let's work through this issue step by step. You mentioned your Fortran (.f90) code can read data from /Users/rob/Desktop/DISSERTATION/DSMCDATA/ without problems, but when you export modified data to the same directory under a new filename, the subsequent verification import fails. Here are the most common culprits and actionable fixes to check:

• Always use absolute paths for exports: If your code lives in /Users/rob/Desktop/DISSERTATION/DSMCCODE/, relative paths might accidentally write files to your code directory instead of DSMCDATA. Explicitly define the full path when opening the export file:

  integer :: export_unit
  open(unit=export_unit, file='/Users/rob/Desktop/DISSERTATION/DSMCDATA/your_new_file.dat', &
       status='replace', action='write', iostat=ios)
  

• Check directory permissions: Run ls -ld /Users/rob/Desktop/DISSERTATION/DSMCDATA/ in your terminal to confirm your user account has write access (look for the w flag in the permission string). If not, adjust permissions with chmod u+w /Users/rob/Desktop/DISSERTATION/DSMCDATA/.

A mismatch between how you read and write data is a super common issue:

• Text files: Ensure the export uses the same column separators, line endings, and data formatting as the original input. For example, if your input uses space-separated values, don’t switch to commas when exporting. Manually open the exported file to spot obvious formatting errors.
• Binary files: Double-check that variable types, array dimensions, and byte order match exactly. If your input uses real(8) (double-precision), don’t write the export as real(4) (single-precision). Even a tiny mismatch will break verification.

• Always close files after use: If you don’t call close(unit=export_unit) after writing, the data might stay in memory buffers instead of being flushed to disk—making it invisible to your import step.
• Avoid unit number conflicts: Reusing the same unit number for read/write operations without closing first can cause crashes or silent failures. Use Fortran 2008’s newunit keyword to auto-assign safe, unique units:

  integer :: import_unit, export_unit, ios
  ! Import step
  open(newunit=import_unit, file='/Users/rob/Desktop/DISSERTATION/DSMCDATA/input.dat', &
       status='old', action='read', iostat=ios)
  if (ios /= 0) then
      print *, "Failed to open input file! Iostat code: ", ios
      stop
  end if
  ! ... read data ...
  close(import_unit)
  
  ! Export step
  open(newunit=export_unit, file='/Users/rob/Desktop/DISSERTATION/DSMCDATA/output.dat', &
       status='replace', action='write', iostat=ios)
  if (ios /= 0) then
      print *, "Failed to open export file! Iostat code: ", ios
      stop
  end if
  ! ... write data ...
  close(export_unit)
  

• Use iostat for error checking: Adding iostat to your open calls gives you concrete error codes to diagnose issues (e.g., code 13 = permission denied, code 2 = file not found).

• Print the exact file paths your code is using for export/import—typos (like missing slashes, capitalization errors) are easy to miss.
• For text files, use the diff command in terminal to compare a small subset of the original and exported data: diff /Users/rob/Desktop/DISSERTATION/DSMCDATA/input.dat /Users/rob/Desktop/DISSERTATION/DSMCDATA/output.dat | head -20

内容的提问来源于stack exchange，提问作者Rob A.