* apply documentation changes to single_include
* retain scope when including a template
* Use data from current scope when rendering an included template.
This allows included templates to access loop variables, which
was not possible before.
* Add test
* throw exception if a file cannot be opened
* Add a new function in utils.hpp: open_file_or_throw: This function
returns an opened std::ifstream or throws by calling `inja_throw`.
* Use this function in Parser::load_file which previously returned an
empty string if the file couldn't be opened.
* Use this function in Environment::load_json which previously threw
a `nlohmann::detail::parse_error` if the file couldn't be opened.
* In Parser::parse_statement: When including files through `include`,
do not attempt to (re-)parse templates from files that were already
included. Additionally, this prevents inja from attempting to load
in-memory templates by their name from disk.
* Add tests that check if an exception is thrown when attempting to
open files that do not exist.
* cmake: enable C++11
* cmake: require C++11 when depending on single_inja
* code style
Commit a5862a0 made the linking step in update_loop_data unecessary
because the loop's copy of the data was made to come from the parent
loop's data rather than the original client's data.
While at it, also add one more nested loop test case.
* Fix for issue #82 - Nested loops
I made the following changes in renderer.hpp
- Removed the member LoopLevel::it - it was not being used and was
causing issues.
- Added LoopLevel::loop_type with a matching enum class to mark the loop
as looping on either a Map or an Array. This was to replace the hard to
understand test for key_name.empty().
- Modified update_loop_data to get the correct data for map type loops
when copying in outer loop control data.
- Modified update_loop_data to only copy in the outer loop data during
StartLoop, it does not need to be done at EndLoop and can be expensive
since it can copy a large part of the json tree.
All the tests pass and the test cases from the issue also now work. Both
GCC and Clang will compile and run the tests with no issue. No issue is
seen with address_sanitizer in either compiler.
* Remove forgotten debugging code.
* Fix for issue #80
When updating the `opening_chars`, line_statment was forgotten. So, a
change to any of the openers would cause line statements to start being
unrecognized.
I added a unit test to cover this case.
* fix indentation
* add update to single_include
