highlighting Hebrew pronominal suffixes via a filter

There's nothing in the way the data is structured that prevents highlighting one lexical segment of a longer grapheme - we do that in search results views. I'm guessing that the issue is one of two things: 1) the developers were just using standard word-breaking characters to define the limits of any visual filter element (or relying on the way surface forms are indexed without reference to the segmentation of a surface form into one or more lexical elements), and/or 2) they tried using the same code we use for search results but were unhappy with the number of dotted circles that imposes in the Hebrew text in the numerous instances where a pronominal suffix begins with a vowel. If you try to highlight or put some other formatting element a consonant without the vowel mark beneath it, or vice-versa, you get at least one dotted circle in the text (more, if other marks like accents are present at the intersecting spot).

A database could be structured to be less fastidious about where to draw the line between one lexical segment and its suffix as a concession to what displays well on a computer screen in conjunction with different highlighting elements. That would make certain searches more difficult to construct (for example if one wanted to find places where the 1cs accusative pronoun is -ani versus -eni versus -ni - they'd all look like -ni without constructing a search to look at the vowel that ends the previous segment). The route the Westminster database takes of including the pronominal suffix with the word it is attached to in one segment avoids most* of these display issues, but isn't as useful for things like reverse interlinears or syntax graphs where these elements really need to be separated out. Perhaps one could write extra code into the way search results and visual filters are handled to prevent orphaned combining marks in the display without changing the underlying segmentation, but I don't know what the performance cost of that would be (and that could introduce some strangeness to the search results in the analysis view where one searches on -ani and gets a list of -ni hits back that happen to be correct, but require looking at the previous segment to understand that the hits are good - though I find dotted circles least disturbing in the analysis view where the segments are broken apart anyway, so the answer there could be not applying the visual filter 'fix' to that particular search results view). (*The one case that would always insert circles is when an article segment is only a vowel letter underneath a preposition, but I think few people would be making visual filters for elements as common as articles and prepositions, so we can probably live with that regardless.) We can kick some ideas around here.