ndb | Plan 9 NDB implementation in Go | Storage library

What is your default character set? It's utf8mb4 by default in MySQL 8.0, which means you count 4 bytes per character.

Therefore your row size is 14,720 characters * 4 bytes/character = 58,880 bytes.

utf8 is a variable-length encoding, so it's likely not every character will be a 4-byte character, but MySQL must assume that any or all characters in these strings might be 4-byte characters. Therefore it has to accommodate the maximum possible bytes per character when calculating the row size.

This issue seems to have been resolved with Google Cloud SDK version 371

On my debian based system i fixed it by downgrading the app-engine-python component to the previous version

I think I've finally figured it out.

From Browsershot's Github page, it recommends executing these commands to install puppeteer.

None of those things should be "wrapped" "in" "double" "quotes." Perhaps try something like this:

If that's the model (a single model) you have and your '''various filters' are just different ways to filter that single model, I don't believe you can sum directly in the queries.

For the Python code itself, you can use the inbuilt sum function and so your code could be something like

In general @gaefan is correct, you can't do this properly.

There's only 3 ways I can see to kind of make this work.

1. the parent name has to also be the key for parent, then you could order Child.query().order(Child.parent).fetch(). The downside is that changing the name of parent is not really possible. You'd have to change its key and all foreign keys on all Child models and other models that had a foreign key to it

2. denormalize parent name as a field on Child, like this:

   class Child(ndb.Model): parent = ndb.KeyProperty(kind=Parent, indexed=True) name = ndb.StringProperty() parent_name = ndb.StringProperty()

Then you could do Child.query().order(Child.parent_name).fetch(). The downside is again that if you want to change Parent.name, you'd have to then update all of it's children.

1. If this is for something like a http handler and you need to return a page of 10 results or something, you could first fetch 10 parents ordered by name, then run a query for all children who have one of those parents Child.query(namespace="test").filter(Child.parent.IN(order_parent_keys)).fetch() (this is what you suggested in your comment). You would just then have to order the children after the fact by parent name. The issue with this method, is this only works for a limited number of parents at a time (I forget how many values you can pass to .IN()). Also the pagination would get weird when you run out of children for the original group of order_parent_keys and need to get the next set.

The right solution for you really depends on your data. How many parents do you have, how many children do you expect each parent to have. Also, how much data you are trying to process, and how fast you have to process it.

For example if you need to process your all or most of your db once a day, Google Dataflow would be perfect for this.

Main select clause:

You can use a regex and a non-regex approach here:

You can use