add support for trim_blocks, lstrip_blocks about jinjava

[jaredstehler]

Implemented as two new parameters on JinjavaConfig... you can enable like:

JinjavaConfig.newBuilder()
  .withLstripBlocks(true)
  .withTrimBlocks(true)
  .build()

206ca3d

[cosmin]

I looked into this a bit more, and I think it's going to be interesting to implement this in this project. Jinja2 itself accomplishes this because it uses regular expressions to parse blocks, and it modifies the regular expression for the start and end of a block based on the values of trim_blocks and lstrip_blocks.
Meanwhile jinjava does this using a scanner, which I think makes this harder to accomplish.

In theory trim_blocks could be easily supported by continuing to swallow a newline following the end of a block, but even that is tricky, because you only want blocks {% ... %} to get this behavior, and not expressions {{ .. }}, so that alone will requires some work.

However lstrip_blocks becomes more interesting. As far as I can tell, the TextNode will get all of the whitespace prefixing a block, so there is not much we can do during parsing, without having to either look ahead and see if there will be a block starting on this line, or by going back and altering the text node preceding a block.

In either case I think this will require some work to get right. However, it's still a very critical feature when using this project for whitespace sensitive items. For example, Ansible uses Jinja2 to parse its templates, and it has trim_blocks enabled by default. There is also a feature request to turn on lstrip_blocks by default because of all of the issues introduced by not having that turned on. So clearly these are important features at least for some use cases, although I can understand that it probably doesn't apply to what HubSpot needs.

[jaredstehler]

Agreed that this sounds like a tricky one to implement, but we should be able to come up with a comprehensive test suite for it. I should hopefully get a chance to work on it this week.