How-To

Recipes for one task each, with the command line, Rust, Python and PowerShell side by side where each can do it. New to trex: start with the tutorial.

• Build a pattern once and reuse it - mark the fields in one line, save the pattern, read new lines with it.
• Query the records of a log - keep entries of several lines that hold, or lack, the patterns you name.
• Count and rank matches - group by a rendered key and total a typed value per key.
• Redact a log and follow it - mask, keep what identifies a row, mask what the log gains.
• Search a tree of files - walk a directory, keep files by glob, list the matching files.
• Index a tree - skip the files a pattern cannot match without reading them.
• Extract fields - typed values, key and value pairs, the columns of a table.
• Lint with rules and fix - findings, CI annotations and fixes.
• Test a pattern file - the texts each declaration must accept and reject.
• Filter by typed values - durations, sizes, addresses and numbers by value.
• Scan many patterns at once - one read, each match under its pattern.
• Explain why a pattern matched - kinds, checks, axis readings and the route.
• Match inside encoded content - JWT headers and claims, base64 contents.
• Summarize a log by templates - the line shapes a log repeats, and their lines.
• Scan input in pieces - matches settled chunk by chunk.
• Rename or reshape matches - templates, typed slices, computed replacements, files.
• Declare your own atoms - your identifiers as tokens: test, find, count, mask.
• Find tables - tables by their repeating token shape, in text and in a tree.
• Segment without delimiters - cut where the text stops predicting itself.
• Find outliers - values out of scale with their window, column or key.
• Match up to a symmetry - case, word shape, subnet.
• Parse with a token grammar - rules over tokens, precedence, ambiguity.
• Split run-together words - every tiling over a dictionary.
• Learn a subword tokenizer - byte-pair merges from a corpus.
• Find deep nesting - tokens past a bracket depth.
• Find repeated or new content - first sightings, and what one log has that another lacks.
• Find rare lines and out-of-order times - rare line shapes, timestamps running backward.
• Find where text changes kind - code inside prose, data inside code.
• Pre-check a corpus for a literal - a presence filter answering without a scan.
• Measure how far text compresses - the code length of a context-mixing model.
• Choose a backend - CPU, device, chunks, and the compressor’s backends.