Columns with dot in accessor render empty cells

[pfo-omicsstudio]

Describe the bug
Columns whose accessor contains a dot (e.g. Group.names) are rendered as completely empty, even though the underlying row data contains values for that key.

For example, rows in this format:

{
  "key": "a6bbf955-7e2a-4664-a6ab-ce23d32dc080",
  "sampleName": "Sample1",
  "value": 3.150405115836441,
  "Group.names": "GroupA",
  "Organism": "Homo sapiens (Human)",
  "GeneNamesPrimary": "PIGBOS1"
}

and a column like this for "Group.names":

{
  accessor: "Group.names",
  title: "Group.names",
  sortable: true,
}

causes the column header shows up, but all cells are empty.
It looks like "Group.names" is treated as a path (row.Group.names) instead of a literal key row["Group.names"]), so the value is never read correctly from the row object.

To Reproduce
To Reproduce

Use this minimal repro component:

import { DataTable, type DataTableColumn } from 'mantine-datatable';

type Row = {
  key: string;
  sampleName: string;
  value: number;
  'Group.names': string;
  Organism: string;
  GeneNamesPrimary: string;
};

const records: Row[] = [
  {
    key: '1',
    sampleName: 'Sample1',
    value: 3.150405115836441,
    'Group.names': 'GroupA',
    Organism: 'Homo sapiens (Human)',
    GeneNamesPrimary: 'PIGBOS1',
  },
  {
    key: '2',
    sampleName: 'Sample1',
    value: 3.25393558322,
    'Group.names': 'GroupA',
    Organism: 'Homo sapiens (Human)',
    GeneNamesPrimary: 'CD300H',
  },
  {
    key: '3',
    sampleName: 'Sample1',
    value: 3.713411715413312,
    'Group.names': 'GroupA',
    Organism: 'Homo sapiens (Human)',
    GeneNamesPrimary: 'NBDY',
  },
];

const columns: DataTableColumn<Row>[] = [
  { accessor: 'sampleName', title: 'Sample' },
  { accessor: 'value', title: 'Relative Expression' },
  { accessor: 'Group.names', title: 'Group.names' },
  { accessor: 'Organism', title: 'Organism' },
  { accessor: 'GeneNamesPrimary', title: 'GeneNamesPrimary' },
];

export function DotAccessorBugRepro() {
  return <DataTable columns={columns} records={records} />;
}

Expected behavior
If accessor is a string, and the row object has a property with exactly that key (even if it contains a dot), the table should read and display that value.

At minimum, there should be a way to treat accessors as literal keys rather than paths, or a documented way to escape/opt out of path semantics.

Screenshots

Additional context
In R, spaces in column names gets substituted with dots, which is what led me to discover this issue.
As most of our user data is generated in R, this is a major problem.

[icflorescu]

Hey Patrick,

This was actually a feature backed-in Mantine DataTable since its inception, see column properties and styling; the accessor supports dot-notation for nested objects property drilling (i.e. department.company.name), which is a common pattern in JS.
That's also why the accessors including dots are "humanized" by default in titles (i.e. company.name gets a default title of "Company name".

If (due to data coming from R, in your particular case) you're not happy with the default, I think this is something that should be addressed in user-land. Maybe you could build a simple adapter that maps things like "Group.name" to "groupName" to adhere to JS naming standards?

[pfo-omicsstudio]

Thanks for the swift reply. I understand why dot-notation accessors exist and I agree they are useful. The issue is that in my case the data does not come from my code - it is user-uploaded data, and most users generate that data in R, which automatically inserts dots into column names.

Because of that:

• I cannot rename or normalize the keys ahead of time (renaming columns = augmenting users' data, which is a big no-no).
• I don't know the column names beforehand - they're dynamic.
• Adding a custom adapter for every table is brittle and easy to miss.

So this isn't about preferring a different naming style; it's that user data legitimately contains dots, and I need a way to treat accessors as literal keys, not paths. Right now there's no clean way to handle user data with dots in column names.