CAT-Tools/DéjàVu X/Assemble


 * Being able to manually assign a degree of preference to TDB entries for Assembling (Tanja Güllicher)

 Reinstate DV3 option to discard portions less than two words long, three words long, etc. from Assemble and AutoSearch box (user definable) (David Turner)

 Insert an exact match or best fuzzy match on control/alt+down as a default option even if AutoAssemble is turned off (David Turner)

 Assemble option to use only lexicon and/or term base entries (disabling use of memory portions for AA) (David Turner)

 Propagate exact matches exactly without trying to adjust new target to match source punctuation (David Turner)

 Consider a propagation with a number change as a fuzzy propagation rather than an exact propagation, by just changing the number and not trying to readjust the target to match the source punctuation (if n°1 = No. 1, n°2 should be propagated as No. 2 and not No2 as currently the case (David Turner)

 Yes, agree to David: don't try to match source punctation if you cannot handle it anyway; this is really annoying!! If DV does want to handle source punctuation, it should be able to know the differences: e.g. it does have a rule that English floating point numbers such as "4.3" should be "4,3" in German, which makes sense most of the time. Then why not give it rules on other types of punctation? Like, English “ (opening double quote) should be „ in German, and ” (closing qu.) should be “, in turn? Could even be made user-definable (per language). (Florian v. Savigny)

 Propagation is of limited use (and the difference of this feature to simple matching is not impressive) if you go on to modify the "mother segment" after propagation has taken place. Exact propagation would be a real feature if the "child segments" (i.e. the ones where the content was propagated) were simply mirroring the "mother segment", i.e. whatever happens in the mother segment happens in all "child segments" simultaneously. Also, one would need navigation for jumping from the "child segments" to their "mother segment" (and back, as in a browser). Or, alternatively, whatever happens in any of a group of identical cells (including the "mother" and all "children") happens in all others simultaneously. It is a behaviour I usually try to emulate manually using Filter for Selection (but this is certainly more error-prone). (Florian v. Savigny)

Fuzzy propagation should not exist, except perhaps in the sense David has suggested above, when only numbers and/or codes differ (though I'm actually unsure of whether it should be called "fuzzy" then). Fuzzy propagation has no advantage over plain vanilla fuzzy matching; rather, it is more confusing and error-prone (plain vanilla matching – and editing – happens during normal editing, i.e. going through the text in reading order, while propagation means jumping unknown distances and "rushing by"). At the very least, it should be possible to disable fuzzy propagation. Or, at the very very least, you shouldn't lose focus when you cancel it. (Florian v. Savigny)