pcstout/csv-schema-py

csv-schema-py's Issues

Add some optional constraints to columns

It would be nice to have the ability to:

• not authorize additional columns (like the additionalItems of json-schema)
• force column order

With these two options, it even becomes possible to omit the column names, as only their places in the row can be used.

For instance:

{
"additional_columns": false,
"column_order": true
}

Optional columns fail data validation

When a non-required column is not present, header validation succeeds but data validation fails with a KeyError. This is because the validator accesses the dictionary entry for the column whether or not the column is actually present.

The patch is fairly simple, and I am setting up a pull request for you. I would appreciate if you can accept and deploy a new version.

Accept custom dialects

First of all, thanks for this package it’s great!

My problem is that the validator only accepts “classic” CSV with a header, a comma as separator, a newline with \n, etc.

It would be really nice to manage other formats, and even nicer to make these formats part of the schema.

To manage formats, here are some ideas:

• Use the Dialects of the native csv library.
• In this line which reads the CSV file, give a custom dialect to DictReader.
• The custom dialect could either be given as an argument of ValidateCsv.__init__, or could be generated from the schema itself. In that last case, a new method generate_dialect could parse the schema and build the dialect.

Example of schema with maximum complexity for the dialect:

{
  "name": null,
  "description": null,
  "filename": {
    "regex": null
  },
  "dialect": {
    "delimiter": "\t",
    "skipinitialspace": true,
    "lineterminator": "\n",
    "quoting": "QUOTE_NONE",
    "quotechar": "'",
    "escapechar": "\\",
    "doublequote": false
  }
  "columns": []
}

I’ll try to do a PR if I can spare some time.

Make some attributes optional

Maybe it was intentional that attributes are mandatory. But I think some of them can have a default value to null.

If one does not want to check the filename, one does not necessarily want to write the whole filename object in my schema.

min and max could also have a default value equal to null.

I think the following attributes must be mandatory:

• name (for the main object and colum objects)
• type (for colum objects)
• columns (for the main object)