CsvProvider has no option to write dates without timezone information.

[f4hy]

The CsvProvider can give inconstant results if one does not need exact times, but just dates and the file is written and read in a different timezone.

type DateProbmes = CsvProvider<"Name, Date\n myname, 2017/09/19">

If you write such a file in UTC timezone, it will be written to a file as

myname,2017-09-19T00:00:00.0000000-00:00

However if this file is then read somewhere in the US, it will give you some local time on 2017-09-18 !!

When reading the csv file, without looking at the file some other way, there will be no way to know what timezone the file was written in so using just objects returned using CsvProvider you can't know if you want the date 09-18 or 09-19.

If the CsvProvider had a method to write the file with only 2017/09/19 then it would be read as the correct day in all local timezones.

[github-actions]

🤖 This is an automated response from Repo Assist.

This is a real issue with CSV round-trips across timezones. Here are the options available today:

Option 1: Use the "date" type annotation (no time component)

If you only need the date portion, annotate the column as "date" to get System.DateOnly, which has no timezone concept at all:

type Report = CsvProvider<"data.csv", Schema="Name (string), Date (date)">
// Date column is System.DateOnly — no timezone, no UTC offset on save

Option 2: Use "datetimeoffset" for explicit timezone tracking

If you need full date+time, annotate as "datetimeoffset" to use DateTimeOffset. This preserves the offset explicitly, making round-trips predictable:

type Report = CsvProvider<"data.csv", Schema="Name (string), Date (datetimeoffset)">

Option 3: Current workaround

The workaround you found — constructing a new DateTime(old.Year, old.Month, old.Day) — is effectively getting DateTimeKind.Unspecified, which is what option 1 does more cleanly via DateOnly.

Potential enhancement

A PreferDateTimeOffset parameter (similar to the recently-added PreferFloats) could make DateTimeOffset the default for all inferred datetime columns, avoiding this class of issue without requiring per-column annotations. Contributions welcome!

Generated by Repo Assist

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@69c1ee19e39e6aaba35519aafefb5cbf314de323

[github-actions]

🤖 This is an automated response from Repo Assist.

I've submitted a draft PR adding a PreferDateTimeOffset static parameter to CsvProvider, JsonProvider, and XmlProvider. This addresses both the reading and context aspects of this issue.

When PreferDateTimeOffset=true, date-time columns (including those without explicit timezone info like "2023-06-15") are inferred as DateTimeOffset instead of DateTime. When writing back with Row.JsonValue.ToString() or CSV serialization, DateTimeOffset values include the offset, making round-trips timezone-aware.

type Readings = CsvProvider<"readings.csv", PreferDateTimeOffset=true>
// date columns → DateTimeOffset instead of DateTime
let row = Readings.GetSample().Rows |> Seq.head
printfn "%O" row.Timestamp // DateTimeOffset with local offset preserved

See the linked PR for details.

Generated by Repo Assist

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@69c1ee19e39e6aaba35519aafefb5cbf314de323

Warning

⚠️ Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• www.google.com