Log cleaner & deduplicator

The Log cleaner folds a log into what actually happened: every line becomes a template — timestamps, ids, hashes, addresses, ports and numbered paths replaced by placeholders — and lines with the same template are counted as one, the first of them kept in full with its stack trace. Filter by level, read the groups by how often they fired, copy the folded log.

Two lines that differ only in what changes from run to run. Timestamps in the ISO, syslog and Apache shapes, bare dates and times, UUIDs, hex hashes and 0x addresses, IPv4 addresses with their ports, ports after a host name, numbered segments of paths, and any number of three digits or more become placeholders; what is left is the template, and equal templates are one group. Words that only look like a hash, and one- or two-digit numbers, stay.

A line that is indented, or starts like a stack frame — at …, Caused by:, … N more, a Python File line — belongs to the line above it. The whole trace is kept once, with the first occurrence of its line, and counts toward that group.

Blank lines, and every repeat after the first of a group — the lines the fold removed. The status line counts them so you can see how much of the log was noise.

Up to 5 MB of text; anything past that is not read, and the page says so. It is read line by line with bounded patterns, so a big dump takes a moment rather than freezing the page.

One entry per group, most frequent first, with the count in front of its first line — × 12 — and the stack trace under it as it was. Levels you switched off are left out.