Validate security records
Separate valid TSV records from rejected rows with physical line references.
A tab is a boundary
Structured logs are useful only when their structure is understood. This exercise uses three tab-separated fields with no embedded tabs or newlines. It is not CSV: quoted commas and escaped line breaks need a real parser, not a quick change of separator.
awk -F '\t' splits on tabs and keeps empty fields. A Bash read loop with a whitespace separator can collapse adjacent separators, which hides missing data. Validate the number of fields before examining their contents. A row with an empty user is different from a row with too many fields.
Reject visibly
Keep accepted data in one file and rejection reasons in another. Physical line numbers include the header and allow a reviewer to locate a rejected record. Define a single precedence for errors so the same row gets the same explanation on every run.
Use anchored patterns for identifiers and exact comparisons for enumerated values. A result containing NOT_OK is not OK. Never treat malformed rows as successful authentication. Reports should be replaced on a rerun; appending old errors would count the same problem twice. The later capstone builds on this separation between usable evidence and evidence limitations.
awk -F '\t' 'NR > 1 {printf "line=%d fields=%d\n", NR, NF}' auth.tsv
Put it into practice.
Validate auth.tsv after its supplied header id<TAB>user<TAB>result. Write valid.tsv with that header and accepted rows in input order. Write errors.tsv with header line<TAB>reason. Reject in this order: NF != 3 as fields; an ID not matching ^[A-Za-z0-9_-]+$ as id; an empty user as user; any result other than FAIL or OK as result. Use physical line numbers. Blank records are invalid. Replace reports on each run.
See an explained solution
export LC_ALL=C
printf 'id\tuser\tresult\n' > valid.tsv
printf 'line\treason\n' > errors.tsv
awk -F '\t' 'NR>1 {
reason=""
if (NF!=3) reason="fields"
else if ($1 !~ /^[A-Za-z0-9_-]+$/) reason="id"
else if ($2=="") reason="user"
else if ($3!="FAIL" && $3!="OK") reason="result"
if (reason!="") printf "%d\t%s\n", NR,reason >> "errors.tsv"
else print $0 >> "valid.tsv"
}' auth.tsv
The field-count check prevents later checks from interpreting missing columns. The ordered validation rules preserve a reproducible rejection reason, and headers are rewritten so repeated runs do not duplicate results.