Simple Templating for CSV->*

These slides are at:

http://www.w3.org/2014/Talks/1028-CSVW-IH/

Goal

• Definition of a simple template language
• Based on Mustache, just because some syntax had to be chosen:-)
• Should not be overly complicated to implement
• Should be obvlivious to target syntax
• Should cover a large enough family of use case

Structure of a template

... "table level" templates
{{#rows}}
... "row level" templates
{{/rows}}

• The template syntax is identical
• The template within {{#rows}} and {{/rows}} is understood to be repeated for all rows

A template syntax

• Simplest template:

{{name}}

• Template with filters:

{{name.filter1.filter2(arg1,arg2)...}}

What it means 1: table level

{{name}}

• If {{name}} is a table level metadata key whose value is a string, replace it with the value
• Otherwise replace it with an empty string

What it means 2: row level

{{name}}

• If {{name}} is a column name in the metadata document, replace it with the value of the corresponding cell
• Otherwise, consider it as a table level template

What it means 3: filters

{{name.filter1.filter2(arg1,arg2)...}}

• Take {{name}} as defined
• Use value as an input to filter1 which is supposed to produce a string value
• Go until the end of all the filters; the output of the last filter is the replacement value
• arg1, arg2 are strings, not templates
• A filter may return undefined, which means breaking the chain with an empty result

Typical filters

upper lower row_number replace(regexp, value) concat(str) preconcat(str) URI_encode datatype language parsed_cell ...

Examples

@prefix : <http://example.org/{{@id}}#> .

{{#rows}}
{ "row"       : "{{.row_number}}",
  "GID"       : "{{GID.number}}",
  "On Street" : "{{on-street}}"
}, 
{{/rows}}

:Inventory%20Date 
   "{{inv-date.parse_cell}}"
      {{inv-date.datatype.preconcat("^^xsd:")}};

Implementation

• I have implemented a jQuery extension that:
  1. Gathers the metadata following the metadata document (except for packaging)
  2. Looks for an extra "templates" key in the metadata
  3. If template is found for the required format, extracts the template, executes it, and outputs the result
  4. Otherwise, implements the direct mapping

Example

$("div.json").each(function(index) {
  ...
  var request = {url: dataset_url, format: "JSON"};
  $.getCSV(request)
    .done( function(csv_data) {
       ...

Example online

http://iherman.github.io/CSVPlus/test-templateonly.html

Possible additions

• If carried forward, some issues have not been addressed:
• Escape characters in templates
• Getting to table level keys that are not top level: key1/key2
• Using table level keys whose values are not strings
• Separate template for a "last" (data) row (e.g., JSON syntax is very sensitive…)
• Inclusion of URI template (the RFC)
• Template to use the name verbatim to use with filters

[] :{{on-street.capitalize}} "value" ;

But no new features should be added, unless it is really necessary!

It is on Github:

https://github.com/iherman/CSVPlus