Compare with Wikibase tabular data model

[nichtich]

fixes #990

[peterdesmet]

Good comparison. Cf. CSVW https://datapackage.org/guides/csvw-data-package/ I would write it a bit more from the viewpoint of someone that is familiar with MediaWiki Tabular Data (and discovers Data Package). Tables could then have 3 columns:

• MediaWiki property I'm familiar with
• Is it supported in Data Package
• Details on how it is supported in Data Package

We could also do it differently (and update CSVW), but I would prefer to align how we write comparison guides.

[peterdesmet]

https://datapackage.org/guides/csvw-data-package/#columns compares properties for users that know CSVW, but want to know if it is supported in Data Package. I wonder if the same approach would make sense here? It would allow to combine the table with the differences.

For example (see source for markdown):

MediaWiki Tabular Data property	Data Package support	Details
name (implied from page)*	Yes	As resource.name
description	Yes	As resource.description, but without support for localized strings
data*	Yes	MediaWiki always requires an array of arrays, while Data Package supports other formats as well as data files referred to by path
license	Yes	As resource.licenses with a name (e.g. CC0-1.0)
schema*	Yes	As resource.schema, see further
sources	Yes	As resource.sources, but without support for Wiki markup

[peterdesmet]

If possible, I would also link the MediaWiki Tabular Data property

[peterdesmet]

Similar as suggestion above, we could express table cf. https://datapackage.org/guides/csvw-data-package/#data-types

---

MediaWiki Tabular Data data type	Data Package support	Details
number	Yes	As number field type, which has more extensive support
boolean	Yes	As boolean field type
string	Yes	As string field type, but without the 400 character limitation and can include \n and \t
localized	No

Missing values in MediaWiki Tabular Data are expressed as null, while in Data Package you need to explicitly list values that should be considered missing in schema.missingValues.

[nichtich]

I'd keep it as a list:

• number subset of Table Schema number (no NaN, INF, or -INF)
• boolean same as Table Schema boolean
• string subset of Table Schema string (limited to 400 characters at most and must not include \n or \t)
• localized refers to an object that maps language codes to strings with same limitations as string type.
  This type is not supported in Table Schema.

Indiviual values in a MediaWiki Tabular Data table can always be null, while in Table Schema you need to explicitly list values that should be considered missing in schema.missingValues.

[peterdesmet]

Cf. suggestions above

MediaWiki Tabular Data field property	Data Package support	Details
name*	Yes	As field.name
type*	Yes	As field.type
title	Yes	As field.title, but without support for localized strings

[nichtich]

changed to:

MediaWiki Tabular Data Property	Data Package Table Schema
name (required) must be a string matching ^[a-zA-Z_][a-zA-Z_0-9]*	name (required) can be any string
type (required) is one of the Data Types above	type (optional) with different data types
title (optional) is a localized string	title (optional) is a plain string

[nichtich]

Thanks, I'll adjust the commit but try to make it usable for both familiar with MediaWiki and Data Package, so there is no "supported yes or no".