openRead returns one more layer of []

final data = await file
        .openRead()
        .transform(utf8.decoder)
        .transform(csv.decoder)
        .toList()

Upgrade from ^6.0.0 to csv: ^7.2.0, openRead returns one more layer of [], and now to read the header, it needs to be data.first.first. Is it working as expected or is it a bug?

[
    [
        [
            "Amount",
            "link-one",
            "Name",
            "Status",
            "Tags",
            "Done",
            "Done (文本)",
            "Created Time",
            "单向",
            "双向1",
            "双向1A",
            "双向1B"
        ],
        [
            "1.55",
            "1.55",
            "HelloKitty ",
            "Pending",
            "Archive, Important",
            "true",
            "true",
            "2026/02/25 11:30",
            "Beta Value 16",
            "Beta Value 16, Upsilon Entry 35",
            "HelloKitty , Okay ",
            "1.55"
        ],
        [
            "2.23",
            "2.23",
            "Okay ",
            "In Progress",
            "Important",
            "true",
            "true",
            "2026/02/26 11:30",
            "",
            "",
            "",
            "1.55"
        ],
        [
            "3",
            "3, 40",
            "Epsilon Data 29",
            "Completed",
            "Archive",
            "true",
            "true",
            "2026/02/27 11:30",
            "Eta Data 90",
            "Omicron Data 92",
            "Sigma Entry 95",
            ""
        ],
        [
            "40",
            "",
            "Sigma Entry 95",
            "Pending",
            "Archive",
            "false",
            "null",
            "2026/02/28 11:30",
            "",
            "",
            "Iota Value 73",
            "3"
        ],
        [
            "5",
            "",
            "Iota Value 73",
            "Completed",
            "Urgent, Review, Important",
            "false",
            "null",
            "2026/03/01 11:30",
            "",
            "",
            "",
            "40"
        ],
        [
            "672.0",
            "",
            "Delta Item 86",
            "Cancelled",
            "Archive",
            "true",
            "true",
            "2026/03/02 11:30",
            "",
            "",
            "Delta Item 86",
            "672.0"
        ],
        [
            "461.0",
            "",
            "Upsilon Data 11",
            "Completed",
            "Urgent",
            "false",
            "null",
            "2026/03/03 11:30",
            "",
            "",
            "Sigma Data 36",
            ""
        ],
        [
            "517.0",
            "",
            "Upsilon Value 90",
            "Cancelled",
            "Archive",
            "true",
            "true",
            "2026/03/04 11:30",
            "",
            "",
            "",
            ""
        ],
        [
            "969.0",
            "",
            "Sigma Data 36",
            "In Progress",
            "Important, Archive, Urgent",
            "true",
            "true",
            "2026/03/05 11:30",
            "",
            "",
            "",
            "461.0"
        ],
        [
            "911.0",
            "",
            "Nu Value 53",
            "In Progress",
            "Important, Urgent",
            "true",
            "true",
            "2026/03/06 11:30",
            "",
            "",
            "",
            ""
        ]
    ]
]

Yeah, that's bothering me too. I rolled back to version 6. Were you able to fix it?

The additional layering is necessary to be a codec.
I remember now, that I had this exact situation already and removed codec compatibility in commit 235d89...

I will either create a new major version with a similar fix or add additional functions:

class CsvCodec extends Codec<List<List<dynamic>>, String> {
  @override
  CsvEncoder get encoder => _encoder;
  
  @override
  CsvDecoder get decoder => _decoder;
  
  /// Stream transformer that emits individual rows.
  /// Use this with Stream.transform() instead of [decoder].
  StreamTransformer<String, List<dynamic>> get rowDecoder => ...;
  
  /// Stream transformer that accepts individual rows.
  StreamTransformer<List<dynamic>, String> get rowEncoder => ...;
}

I might then even deprecate the current functions.

Not sure yet. In any case expect a fix in the next days.

Thanks for reporting.

Navigation Menu

Repository navigation

openRead returns one more layer of [] #77

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

openRead returns one more layer of [] #77

Description

Activity

IlyaOstashkov commented on Mar 15, 2026

close2 commented on Mar 16, 2026

close2 commented on Mar 19, 2026

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions