Note

The following relates to memory management.

For the most part, users can specify the image to be OCR'ed by name
and get the results with getText(), getConfidences(), getBoxes(), getAlternatives().

Alternatively, it can be useful to create a TessBaseAPI object via the tesseract()
function and to use this to query additional information and set variables, etc.

However, we get the results from a ResultIterator obtained from a TessBaseAPI.
It is natural to expose this to end users for more advanced use.
However, one can get an iterator as an R object and then discard the associated tesseract object.
The tesseract object will be garbage collected at some point, and then the iterator is then invalid 
and one can cause a segmentation fault.

Similarly, it can be convenient to get the current image being processed by the tesseract object.
Again, however, if we retrieve this as an R object and the tesseract object is garbage collected,
the image is invalid and using it will cause problems.

We use the ResultIterator internally where we can ensure that the tesseract object is not
garbage collected when we have a reference to the iterator.

We could develop a more elaborate mechanism to track ResultIterator and Pix objects
that are queried from a given tesseract object. However, for now, it doesn't seem
to be worth the complexity.


The basic sequence of events leading to an invalid iterator is 

Time ->

tesseract  ->  iterator  -> GC tesseract -> use invalid iterator