W3C home > Mailing lists > Public > public-csv-wg@w3.org > June 2014

Re: How to distinguish field (cell) references in JSON, and other questions about templates

From: Gregg Kellogg <gregg@greggkellogg.net>
Date: Wed, 11 Jun 2014 08:06:26 -0700
Message-Id: <461672A2-B9D2-4560-8155-9A225DEFE2C0@greggkellogg.net>
Cc: "public-csv-wg@w3.org" <public-csv-wg@w3.org>, Andy Seaborne <andy@apache.org>
To: "Tandy, Jeremy" <jeremy.tandy@metoffice.gov.uk>
On Jun 11, 2014, at 3:17 AM, "Tandy, Jeremy" <jeremy.tandy@metoffice.gov.uk> wrote:
> 
> Hi - having read through the [Generating RDF from Tabular Data on the Web][1] doc, I wonder how the [field references][2] work in a JSON template ...
> 
> (aside: should "field references" now be "cell references" given last week's agreement on terminology?)
> 
> The field reference is a column (short) name enclosed in braces; but JSON also makes extensive use of braces. How would an application reading a JSON template distinguish a field reference from normal JSON?

I think this was captured on the call today; the template mechanism works over text, rather than JSON, so the fact that JSON also uses curly braces does not factor in. Basically, a template has a possible prefix (e.g., '=') followed by a string that MUST denote a column heading, followed by an optional modifier starting with a colon (':'), all enclosed in curly bracess. Note that this can't match valid JSON, where the curly brace will always precede a double quote ('"') with optional white space.

> Perhaps a the "template" definition in the CSV metadata should specify the character used to enclose field references?

Unnecessary complication, IMO. Plus, we're simply leveraging the URI Template RDF, which uses curly braces.

> Assuming the JSON syntax from the [Metadata Vocabulary][3] doc e.g. 
> 
> ...
>  "template ": {
>    "name": "My JSON transformation template",
>    "source": "my-json-template.json",
>    "hasFormat": "application/json",
>    "description": "blah blah blah",
>    "type": "template",
>    "cell-ref-enclosure": ["%%","%%"]
>  }
> ...
> 
> The keys are taken from the list in the Metadata Vocabulary (these are from Dublin Core Terms), with the exception of "cell-ref-enclosure" which I've used to specify the "cell reference escape character sequence"; I think there must be a better name! 
> 
> The value must be an array of length 2; the start and finish character might not be the same: e.g. the default brace enclosure would be:
> ...
>    "cell-ref-enclosure": ["{","}"]
> ...
> 
> Given the escape sequence "%%" & "%%"; the JSON template might then include statements like:
> 
> ...
>  "result": {
>      "def-op:airTemperature_C": { "value": "%%air-temp%%" },
>      "def-op:dewPointTemperature_C": { "value": "%%dew-point%%" }
>  }
> ...
> 
> You'll also notice that I've assumed that the template is not _inlined_ in the CSV metadata document; I've used (perhaps incorrectly) "source" to refer to the target. 
> 
> It also occurs to me that the template definition in the CSV metadata should specify the Content-Type created by a target transformation - a given CSV metadata doc might include references to multiple transformation templates; and users should be able to ask for _one_ of those by name. I've used "hasFormat" to specify this.

That would be useful, particularly given the templates process simple text strings. We should also describe how the result is treated if it is not valid for the specified content-type.

Gregg

> Jeremy
> 
> [1]: http://w3c.github.io/csvw/csv2rdf/ 
> [2]: http://w3c.github.io/csvw/csv2rdf/#dfn-field-reference
> [3]: http://w3c.github.io/csvw/metadata/index.html 
Received on Wednesday, 11 June 2014 15:06:55 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:21:40 UTC