- Massive Upgrade to Inputs Infrastructure - We've had this cooking for some time, and are excited to upgrade our entire inputs infrastructure. Yes, this is a backend change, so what's the upside for you? Higher throughput of EPS, more stability, better monitoring of issues, and since it vastly improves our upgrade speed you'll see more of our updates even faster.
- Truncatable Varchar - Have you been writing transform code to make sure that your varchars don't overflow their character limit and cause conversion errors? You'll love the new truncatable varchar. Just check the "Truncate" checkbox on a varchar option in the mapper screen, and we'll make sure to truncate the string to the appropriate length. You'll note the type displays as "varchar(length)T", with the T indicating the truncation setting.
- Reduction in COPY Errors - Let's face it - sometimes it's just not possible to copy to Redshift. Whether it's busy, down, or there's an issue with your data, it means stuff ends up in Restream and not where you want it to be. We've made a couple of big improvements there:
- We now retry the COPY command twice more to increase the likelihood of success in the event that Redshift is down or busy.
- We made COPY commands each run in their own transaction, so the failure of one doesn't result in the failure of others. That means that if there's a data issue in a particular table it will only fail the events for that table, instead of an entire batch which could include data from other tables.
- When setting your
metadata.event_typeto the value of a particular field, sometimes that field value is empty or the field doesn't show up at all. In the past this meant the event got mapped to "null" or "undefined", but now we map it to the input_label to make tracking it down and fixing it even easier.
- We now correctly alphabetize capitalized and uncapitalized event types in the mapper.
- Fixed a bug where after discarding an event type it showed as not discarded until the page was refreshed.
- Some events were in danger of not being transformed for a split second when the transform was being updated. They are now safe and will always be transformed.
- We changed the way we calculate varchar/char length to better align with Redshift, now checking length in bytes instead of in characters.
- Fixed a bug where after deleting an S3 input it came right back as an un-editable input. It now stays gone.
- The Restream button was disappearing in some edge cases, so we made sure it was always there and correct.