Hugging Face's logo

Datasets:
Shpigford
/
cron-schedule-conversion

Tasks:
Text Generation
Modalities:
Text
Formats:
parquet
Languages:
English
Size:
1K - 10K
Tags:
cron
systemd
schedule
devops
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
cron-schedule-conversion / README.md
Shpigford's picture
Shpigford
Upload folder using huggingface_hub
2cf90d4 verified
preview code
|
Raw
History Blame Contribute Delete
4.04 kB
metadata 
license: apache-2.0
language:
  - en
size_categories:
  - 1K<n<10K
task_categories:
  - text-generation
tags:
  - cron
  - systemd
  - schedule
  - devops
pretty_name: Cron Schedule Conversion
configs:
  - config_name: default
    data_files:
      - split: train
        path: train.parquet
      - split: validation
        path: validation.parquet
      - split: test
        path: test.parquet

Shpigford/cron-schedule-conversion

Natural-language schedule descriptions paired with cron expressions and systemd OnCalendar strings.

Example

field value
natural_language every Tuesday at 3am except December
cron 0 3 * 1-11 2
systemd Tue *-01..11-* 03:00:00
note Months 1-11 only excludes December.
split test

Schema

column type description
natural_language string The schedule expressed in natural English
cron string Equivalent cron expression (multi-line for compound schedules)
systemd string Equivalent systemd OnCalendar string
note string Caveats, especially when cron cannot exactly express the schedule
split string "train" / "validation" / "test"

Coverage

  • Standard schedules (daily, weekly, monthly, every N minutes/hours)
  • Holidays (Christmas, Thanksgiving, Black Friday, Halloween, Valentine's Day, Independence Day, leap day)
  • Casual / fuzzy time references ("lunchtime", "before bed", "first thing", "crack of dawn", "end of business")
  • Ordinal weekdays ("second Tuesday of the month", "last Friday")
  • Negative specifications ("every day except Sunday", "all months except December")
  • Sub-minute intervals (cron limitations annotated in note)
  • Awkward intervals (every 90 minutes, every 7 minutes — cron limitations annotated)
  • Compound schedules requiring multiple cron lines (each separated by \n in the cron field)
  • systemd-specific features (OnBootSec=, Persistent=, RandomizedDelaySec=)
  • Time zones (UTC, ET, PST, etc — cron sets TZ=, systemd appends timezone)
  • Typos and informal phrasings

How it was built

  1. Hand-crafted seed examples (~85 cases) covering all the trick categories above.
  2. Templated generation — combinatorially expanded with random times, days, and intervals.
  3. LLM paraphrasing — each seed expanded into ~8 different natural-language phrasings via Claude Sonnet, with cron/systemd labels held constant.
  4. LLM novel-category generation — Claude generated new (nl, cron, systemd) tuples for weird categories, then a second pass verified each example for semantic correctness; rejected examples were dropped.
  5. All cron expressions validated with croniter.

Caveats

  • The model trained on this data emits a single best-guess for ambiguous fuzzy times (e.g., "morning" → 7am). Reasonable but opinionated.
  • For schedules cron cannot express natively ("every other week", "last Friday of month"), the dataset uses the closest valid cron and notes the limitation. Don't take cron field at face value when note is non-empty.
  • Vixie cron OR-matches DOM and DOW; the dataset assumes AND-matching. Verify on your specific cron implementation.
  • English only.

License

Apache 2.0.

Citation 

@misc{cron-schedule-conversion,
  author = {Pigford, Josh},
  title = {Cron Schedule Conversion Dataset},
  year = {2026},
  howpublished = {Hugging Face},
  url = {https://huggingface.co/datasets/Shpigford/cron-schedule-conversion}
}