moontick
Audit a fixed-interval sampling plan against observed tick timestamps.
testing
time-series
sample-plan
audit
Author
Version
0.1.1
License
MIT
Repository
Last updated
last month
Downloads
33
#MoonTick
MoonTick audits a declared fixed sampling plan against observed integer-millisecond timestamps. It reports exact grid coverage, compresses missing points into index ranges, and identifies duplicates, out-of-order records, off-grid timestamps, and out-of-range timestamps. An empty input is a valid observation of a completely missing plan.
MoonTick is a MoonBit library and native CLI. This module is version 0.1.1; version 0.1.0 is published on Mooncakes. Competition acceptance has not been claimed.
The install command for version 0.1.1 is moon add z2823253773-p/moontick@0.1.1. In a separate MoonBit project, add "z2823253773-p/moontick/core" to moon.pkg imports, then call @core.make_grid and @core.audit. An independent install check built a fresh project against the published 0.1.0 registry copy and ran complete and missing-point cases.
#Build and try it
Use the pinned development compiler moonc v0.10.14+7d59c7ec9. From this repository:
moon check --target native
moon test --target native
moon build --target native
./_build/native/debug/build/cmd/moontick/moontick.exe check \
examples/synthetic/complete.ticks --start-ms 0 --end-ms 60 --step-ms 15 --format jsonThe last command audits the half-open window [0,60) at 15 ms intervals. Its four expected timestamps are 0,15,30,45; the example covers all four and exits 0. Replace complete.ticks with duplicate-and-missing.ticks to see a duplicate that cannot repair a missing point; that audit exits 1. Omit --format json for the default text report.
Three runnable synthetic examples cover minute archives, aggregated bucket starts, and duplicate records. Each has a complete and a failing input, a declared plan, and expected counts. MoonTick does not ingest the source systems directly; callers supply extracted timestamp ticks.
For a short walkthrough, see the five-minute demo. The independent trial guide gives a new user reproducible steps and a way to report both successful and failed attempts.
The input format is ticks: no header, one canonical signed Int64 millisecond value per line. A zero-byte file is valid and reports the whole plan as missing. Blank lines, BOM, non-ASCII bytes, CSV fields, whitespace around values, and malformed integers are rejected. The declared window and step must form a positive whole number of grid intervals; MoonTick never infers them from the data.
moontick check FILE --start-ms START --end-ms END --step-ms STEP
[--format text|json] [--detail-limit N]--detail-limit defaults to 1000 and accepts 1–10000. It limits displayed details, not complete counts. JSON audit reports use moontick.audit.v1; errors use a separate moontick.error.v1 document. All timestamps, counts, and grid indices in JSON are decimal strings to preserve Int64 precision.
| Exit | Meaning |
|---|---|
| 0 | Audit passed, or help/version requested |
| 1 | Audit completed and found a data problem |
| 2 | Usage, configuration, input, or resource error |
| 3 | File I/O error |
| 4 | Internal error |
The reusable core package exposes make_grid(start_ms, end_ms, step_ms) and audit(times, grid, detail_limit). ticks_input parses the strict line format; report renders text and JSON. The generated pkg.generated.mbti files document the current public signatures. Product code is MoonBit; Python is used only for independent tests.
#Verify a checkout
moon fmt --check
moon check --target native
moon build --target native
moon test --target native
MOONTICK_BIN="$PWD/_build/native/debug/build/cmd/moontick/moontick.exe" python3 tests/cli/test_cli.py
MOONTICK_BIN="$PWD/_build/native/debug/build/cmd/moontick/moontick.exe" python3 tests/oracle/test_differential.pyThe oracle uses enumerated small grids, a fixed random seed, and metamorphic checks rather than the product's gap algorithm. GitHub Actions run 35747020671 passed on macOS arm64 and Linux x86_64 at commit e047c4d with the pinned compiler. Each platform ran 77 MoonBit tests, 42 real-process CLI tests, and the independent oracle. The publisher has not provided a usable checksum for the core archive, so its supply-chain verification remains open.
#Scope and limits
The v0.1 design audits one finite series on one constant-step integer grid. It does not parse multi-column CSV, infer a timezone, tolerate jitter, handle variable calendar intervals, predict missing values, or repair input. Resource limits are 32 MiB of input, 250,000 records, and 20 bytes per token. The expected grid may be much larger because missing intervals are compressed rather than enumerated in the product.
The implementation contract and review evidence are in the public source repository's specification, five-minute demo, and evidence directory; these documents are intentionally excluded from the small Mooncakes archive. Independent third-party user feedback remains pending.
Licensed under MIT.
Install
Download zipAudit a fixed-interval sampling plan against observed tick timestamps.
testingtime-seriessample-planaudit
MIT
Updated last month
33 downloads