Add support for encoding data as a raw image bitmap to UTF grids

[mcuelenaere]

This adds support for getting the raw image data rendered by the
mapnik grid renderer. The data is returned as 64-bit signed integers or
32-bit signed integers (host endian), depending on whether BIGINT
is true or not.

[flippmoke]

I am somewhat confused by this, why not just use the Image object to get raw image data

[mcuelenaere]

@flippmoke How would you access the Image object? AFAICS, there's no way to access the underlying buffer from a mapnik::grid or mapnik::grid_view through the current Python bindings (except perhaps by repeatedly calling the get_pixel() method)

[flippmoke]

My question is more why are you using a grid object rather then an Image object.

https://github.com/mapnik/python-mapnik/blob/master/src/mapnik_image.cpp#L454-L456

tostring might be helpful then for what you are looking to do?

[flippmoke]

Also if you did want the raw character buffer from the grid does that really help because you wouldn't have a mapping to what the character were?

[mcuelenaere]

@flippmoke I experimented with UTF grid, but found its resolution to be too low for our use case and increasing the resolution inflates the JSON too much. So by getting the raw grid data (which is basically an array of 64-bits containing the feature IDs (which can be downscaled to 32-bit, 16-bit or 8-bit; depending on the amount of features)), we want to send these as images down to the client and extract the data to get the corresponding feature ID in JavaScript. A mapping to the actual feature metadata will be provided out of band, eg with a static JSON file.

[flippmoke]

Okay that makes more sense. So some things I want to talk about before we would be able to merge something to help you out.

• I am not sure about the API here, encoding was originally so that we could provide different strings types of the same JSON output. I would rather not change this to raw as it is a different context for the entire method. Therefore, I would suggest this be a different method all together
• Your pull request request doesn't include any tests for this code. One of our goals in all our mapnik repositories is to get to 100% test coverage. I haven't yet setup coveralls for python mapnik, but I do want to see any new code well tested.

If you need any help towards achieving these things let me know please, more then willing to give input or help.

Thanks for your pull request!

[mcuelenaere]

Sure, what API are you suggesting? getRawData()?

I'm off the coming week so I probably won't be responsive.

[flippmoke]

That would be fine, do we need to have a way to retrieve what the raw symbols mean then at this point, etc. If they are not included in the getRawData.

[mcuelenaere]

Aren't the raw symbols the feature IDs? Unless you want to keep that an implementation-specific thing and allow for changing the meaning in the future? (in which case you do need a mapping)