For reference re: justifiable homicide
In the dispute over how many self-defense cases occur, one data point often cited by those seeking to minimize the number is the FBI Uniform Crime Report's count of justifiable homicides (a legal category that includes self-defense). This is usually in the range of 900 a year, including several hundred by police. While that doesn't count self-defense that doesn't result in the perp's death, it is argued that it is inconsistent with hundreds of thousands or millions of defensive uses annually.
What's not realized is that the FBI count is artificially defined in a way that far undercounts defensive uses. The usual definition of self-defense with a deadly weapon is use of force immediately necessary in light of a reasonable belief that the perp is likely to inflict death or serious bodily injury.
But the FBI UCR Reporting Handbook at pp. 17-18 uses a completely different definition. Reporting officials are instructed, in the case of use of force by a non-LEO, to include under justifiable homicide only killings "The killing of a felon, during the commission of a felony, by a private citizen."
The illustration given (do NOT list as justifiable a situation where a citizen shoots a fellow attacking him, in a crime of passion, with a broken bottle -- the author must have watched too many 1950s movies about street fights) makes it apparent that the assault itself does not count as "commission of a felony."