Date ranges (e.g., 2014/2019) are incorrectly parsed as single year in metadata

[nneune]

Before opening an issue, please:

• Make sure you are using the latest version using datasets --version
datasets version: 18.15.0
• Review our documentation

Describe the bug

Date ranges are not correctly downloaded with the datasets CLI. Sequences with collection date ranges in the format YYYY/YYYY (e.g., 2014/2019), which is a valid ENA format, are truncated to only the end year in the downloaded metadata.

For example, sequence MW179421.1 was collected between 2014 and 2019 (shown as 2014/2019 on GenBank), but datasets reports only year=2019 in the metadata file.

This affects all NCBI entries with this date format.

Impact: This is problematic for temporal analyses and phylodynamic inference, as the date uncertainty/range is lost, potentially biasing tip date calibrations and evolutionary rate estimates.

Indicate what operating system you're using

Linux (WSL)

To Reproduce

Steps to reproduce the behavior:

1. Run:

   datasets download virus genome taxon "39054" --no-progressbar --filename data/ncbi_dataset.zip

2. Unzip and open data/ncbi_dataset/data/data_report.jsonl
3. Search for accession MW179421.1 and check collectionDate

Actual output:

"isolate": {
  "collectionDate": "2019",
  "name": "ZY2017-12-EV71"
}

Expected behavior

The date range should be preserved, either as:

• Original format: "collectionDate": "2014/2019"
• Structured fields: "collectionDateStart": "2014", "collectionDateEnd": "2019"

Currently, only the end year (2019) is retained, losing the 5-year uncertainty window.

Note: This issue appears specific to the / format for date ranges. Other range formats have not been tested.

[ericcox1]

Hi @nneune,

Thank you for opening this issue. We are going to look into whether we can add support for the date range.

Best,
Eric

[joverlee521]

Hi @ericcox1,

I noticed this morning that NCBI Dataset is returning date ranges!

For example, MH260471 has /collection_date="2001/2002" and NCBI Datasets is returning "collectionDate":"[2001 TO 2002]". This format doesn't seem to be documented anywhere yet, so just wanted to confirm that this is the expected format for date ranges going forward.