Locate a row with fewer fields

What a width warning measures

Every row in a well-formed table has the same number of fields as its header. When one does not, the inspector counts it under Rows with a different width. That count compares each row with the first row and nothing else. It cannot tell whether a value is missing from the record, the header has an extra column, or the wrong delimiter was chosen for the whole text.

Each of those causes needs a different repair, and the wrong one moves data into the wrong column.

The Columns figure in the summary is the width of the widest row, so a row with too many fields raises it above the header's width. When Columns is larger than the number of header fields, look for a value that was split.

Reading a,b,c over 1,2

The page example is a header a,b,c and a record 1,2, with comma as the Delimiter. The inspector reports 2 rows, 3 columns and 1 row with a different width: row 2 has 2 cells instead of 3. Columns reports the widest row, which here is the header.

The preview shows the row as it was parsed, with nothing added. Which of the three positions is missing is not something the tool can know: 1,2 could be a and b with c empty, or b and c with a missing.

Tracing the cause

• Find the short row in the preview and the same record in the source.
• Compare its fields one by one with the header to see which position is missing.
• If most rows are uneven, the Delimiter is probably wrong; try the alternatives or Detect delimiter.
• If only rows with long descriptions are uneven, check for unquoted commas or quotes inside those values.
• Once you know what the missing position means, add the empty field with a delimiter in the right place in a copy and calculate again.

Repairs to avoid

• Adding delimiters at the end of every short row: that assumes the last column is the empty one.
• Deleting short rows: they are often real records with one blank field.
• Removing a header column to make the counts match, without checking whether some rows use it.
• Assuming the Tablora app's grid has the same problem. The app keeps every row of a table the same width by adding empty cells at the end, so a misplaced value there looks like an empty last cell instead of a short row. Compare such rows with the Source · page view and move the value with a cell edit if needed.

Original synthetic example

Input

a,b,c 1,2

Delimiter

,

Expected records

2

Columns in the first record

3

Width difference

Row 2 has 2 cells

Tools

• delimited-inspector

Sources

• RFC4180: Common Format and MIME Type for CSV Files