Andrew Snare

Results 36 issues of Andrew Snare

This pull request updates the `timestamp()` and `clientTimestamp()` providers to produce native timestamps that can be mapped to Avro fields with the timestamp logical types. Such records are compatible with...

enhancement

A fairly common misconfiguration for browser-based endpoints is having an explicit cookie domain set to a domain which doesn't match the site in which the tracking script is loaded. This...

enhancement

Right now Divolte has some basic support for JavaScript modules, but it doesn't work easily with [Require.js](http://requirejs.org), which seems to be the most popular module loader/injector currently in use. Some...

bug
enhancement

Divolte currently uses Avro to serialise events as Kafka messages. The [serialisation format](https://avro.apache.org/docs/1.8.1/spec.html#Data+Serialization) includes only the event data and is not self-describing. Deserialisation requires knowledge of the write-schema or the...

feature

At present our mapping DSL cannot map to values onto enum Avro fields. Mapping verification during startup detects a mismatch between the mapping and the Avro type. This issue covers...

enhancement

This test is a bit flaky due to a race condition. The pages load the script with defer/async. Although most actions in the test include a wait to ensure the...

bug

At present the javascript doesn't correctly handle `document.currentscript` being `null` during load. (Various modern loader tricks can lead to this, one of which is noted in pull request #179.) This...

bug

While fixing issue #157 it was discovered that there seems to be a bug where events can sometimes stop working on Opera 12.12 on Windows 7. Specifically, some of our...

bug

At present Divolte always does a City-based lookup when mapping GeoIP metadata, even if only less granular country data is going to be mapped. This fails on the Country-Lite database,...

bug
enhancement

This pull request implements support for configuring browser sources to send a 204 (No Content) response instead of the tiny 1x1 image, as discussed in #107. Documentation still needs to...

enhancement