swath show
swath show prints a committed dataset's summary, or an interrupted run's recorded command and progress.
Synopsis
swath show DIR [OPTIONS]Description
For a committed dataset, swath show prints the manifest's summary for people, in reading order: the verdict first, then labelled lines for the source, when it was listed (local time, the zone named), what the dataset holds (its rows, the bytes of the objects they list when every size is known, and the key range, clipped to a terminal's width and left out where it cannot fit), the coverage, and last how it is stored (the Parquet parts, their bytes and the swath schema; -v adds the SeekParquet format id before the schema and the build that wrote the dataset after it):
$ swath show media-inventory
media-inventory: complete
source /home/you/media (whole tree)
listed 2026-11-15 21:03:25 to 21:03:25 CET (0.0s)
contents 12 entries, 101 MiB; keys "archive/" .. "videos/trip.mp4"
coverage nothing unreadable; no filters; no exclusions
storage 1 part, 4.05 KiB of Parquet (schema swath/local-entries/v1)The verdict is complete when nothing is missing: the listing reached the end of its scope and read every entry. Otherwise it is partial: and what is missing: partial: 2 unreadable entries (entries that could not be read; a diagnostic sample says partial: --max-keys reached). The coverage line counts the unreadable entries, names the filters the run applied by their rules (filtered: include=*.jpg, min_size=200k), and what was left out by design (excluded: ..., an output directory inside the tree). A complete dataset without filters covers the source on the line below the verdict, minus its exclusions (Understand your results, The swath dataset format). There is no machine form of this summary: a program reads manifest.json.
For the work directory of a dataset written to a bucket (-o s3://..., gs://... or az://... with --work-dir DIR), it prints the same summary of the manifest the run committed there, under the destination its commit record (DIR/commit.json) names, and a last line saying the commit came from that work directory and whether its cleanup is finished (finalized) or a swath resume DIR still finishes it:
$ swath show work
s3://out/inv/: complete
...
committed from the work directory work (finalized)For an interrupted run, it prints when it started, the command the journal recorded, the parts saved so far and where swath resume will continue:
$ swath show logs-inventory
logs-inventory: interrupted run
started 2026-11-15 21:04:35 CET
command swath list --endpoint-url=http://127.0.0.1:19480 --anonymous --output=logs-inventory --request-rate=10 --part-interval=1s s3://bkt/logs/
saved 1 part, 1,000 rows; swath resume lists after "logs/2026/00/app-00999.log.gz"-v adds the journal's version to the start line and, after the command, the identity a resume must match: the provider, the schema, the writer recipe, the part size and the source's frozen settings, each with its value (a setting without one, such as the endpoint of a run against AWS, is left out):
$ swath show logs-inventory -v
logs-inventory: interrupted run
started 2026-11-15 21:04:35 CET (journal version 1)
command swath list --endpoint-url=http://127.0.0.1:19480 --anonymous --output=logs-inventory --request-rate=10 --part-interval=1s s3://bkt/logs/
identity provider s3, schema swath/s3-objects/v1, writer swath/parquet-r5, part size 64.0 MiB, source s3://bkt/logs/, endpoint http://127.0.0.1:19480, region us-east-1, path style true, fips false, dual-stack false, fetch owner false
saved 1 part, 1,000 rows; swath resume lists after "logs/2026/00/app-00999.log.gz"The command is recorded in its canonical spelling, each option --name=value, then SOURCE. It is printed as a shell line: an argument the shell would split or expand (a path with a space, a leading =) is quoted, in single quotes, or on Windows in double quotes, so the line can be pasted as it is.
swath show reads no rows. swath du --from-dataset DIR totals them per prefix (Total a source); to query a dataset otherwise, read the parts its manifest lists (Query an inventory).
Options
--journal
dump the journal's header and records as JSON lines (diagnostic: the journal's private format, which may change in any release)
swath show also takes -v, -q, --color and --log-format (swath(1)).
Examples
Summarize a dataset:
swath show logs-inventoryDump an interrupted run's journal:
swath show logs-inventory --journal--journal is a diagnostic: it prints the journal, swath's private resume record, as it is stored. Its members, their spelling and their types are not an interface and may change in any release; a program reads the run report, progress.json, error.json and the manifest instead.
Exit status
0 when it printed the summary, or with --journal the journal (an interrupted run's, or the one a committed dataset's run kept); 1 when DIR does not exist, holds neither a dataset nor an interrupted run, or with --journal holds no journal (a committed dataset whose run kept none); 2 for a usage error.
Files
DIR/manifest.json, DIR/.swath/journal; for a bucket dataset's work directory DIR/commit.json and the manifest it committed, under DIR/.swath/.
See also
swath list, swath resume, Query an inventory, The swath dataset format