Globs View

Query a collection of Globs by breakdown and output — the pivot-table shape: group the data by a series of dimensions, aggregate a few measures, and get back a hierarchical tree (or a flat CSV export). The dimensions are not written in code: they are discovered from the GlobType, published as a dictionary, and picked by name in the request. That is what makes it an endpoint an interactive UI can drive.

Requirements

Java 21, org.globsframework:globs, globs-gson; globs-http for the ready-made server.

Installation

<dependency>
    <groupId>org.globsframework</groupId>
    <artifactId>globs-view</artifactId>
    <version>5.1.4</version>
</dependency>

The three steps

1. The dictionary — createDictionary(type) walks the GlobType, descending into GlobFields and GlobArrayFields, and returns a Glob listing every dimension it found: a uniqueName (the one the request uses), the path to reach it, its type, and whether it is flagged IsSensibleData.
2. The request — a ViewRequestType Glob: the ordered breakdowns (the nesting of the result tree), the output measures, and an optional filter.
3. The view — buildView(dictionary, request) returns a ViewBuilder; createView() gives a View you feed Globs into, and toGlob() renders the tree.

ViewEngine viewEngine = new ViewEngineImpl();

Glob dictionary = viewEngine.createDictionary(ViewType1.TYPE);

MutableGlob viewRequest = ViewRequestType.TYPE.instantiate();
Glob[] breakdowns = dictionary.get(DictionaryType.breakdowns);
viewRequest.set(ViewRequestType.breakdowns, new Glob[]{
        br("Name1", breakdowns),
        br("NameSub2", breakdowns),
        br("Name2", breakdowns),
        br("NameSub1", breakdowns)
});
viewRequest.set(ViewRequestType.output, new Glob[]{
        ViewOutput.TYPE.instantiate()
                .set(ViewOutput.uniqueName, br("qty", breakdowns).get(ViewBreakdown.uniqueName))
                .set(ViewOutput.name, "quantity")
});
ViewBuilder viewBuilder = viewEngine.buildView(dictionary, viewRequest);

View view = viewBuilder.createView();
View.Append appender = view.getAppender(ViewType1.TYPE);
globs.forEach(appender::add);
appender.complete();
Glob result = view.toGlob();

Given a list of ViewType1 globs, the result is the breakdown tree, each node carrying its aggregated outputs:

{
  "name": "",
  "nodeName": "root",
  "__children__": [
    {
      "name": "n1",
      "nodeName": "Name1",
      "__children__": [
        {
          "name": "sub21",
          "nodeName": "SUB2.NameSub2",
          "__children__": [
            {
              "name": "n2",
              "nodeName": "Name2",
              "__children__": [
                {
                  "name": "subN1",
                  "nodeName": "SUB1.NameSub1",
                  "output": { "quantity": 1.0 }
                }
              ],
              "output": { "quantity": 1.0 }
            }
          ],
          "output": { "quantity": 1.0 }
        }
      ]
    }
  ]
}

View.getAccepter() answers, per dimension name, whether the current request wants it — so a data source can skip loading a column nobody asked for, before any Glob is built. A view that would expand past globs.view.max.node (500 000 by default) raises TooManyNodeException rather than filling memory.

Filters

ViewRequestType.filter holds a filter tree, again as Globs: EqualType, NotEqualType, ContainsType / NotContainsType, StrictlyGreaterType / GreaterOrEqualType, StrictlyLessType / LessOrEqualType, IsNullType / IsNotNullType, combined with AndFilterType, OrFilterType and NotType, each pointing at a dimension through a PathToField. Being Globs, a filter arrives as JSON from a client, and FilterBuilder compiles it once into a Filter.

Serving it over HTTP

HttpViewServer wires the whole thing onto globs-http, with the OpenAPI description included:

Route	What it does
GET /sources	the data sources a DataAccessor exposes
GET /dictionary	the dimensions available on a source
POST /computeView	takes a ViewRequest, returns the tree — or a CSV file with outputType=csv (leafOnly for the leaves only)

new HttpViewServer("0.0.0.0", 8080, dataAccessor);

Implement DataAccessor / Source to plug in where the Globs come from; CsvExporter is the flat rendering used by the csv output type, and DummyServer is a runnable example.

Fields annotated IsSensibleData are flagged in the dictionary so a server can refuse to break down on them; StringAsDouble lets a string dimension be aggregated as a number.

Building

mvn -o test

License

Apache License 2.0 — see https://www.apache.org/licenses/LICENSE-2.0.txt.

Links

• Globs Framework
• GitHub repository