Printf format specifiers

Question

I've been trying to cobble together the format of printf into a sort of linear format. Is the following a correct understanding of the possible printf formats?

% <justification: [-]?> <sign: [ +]?>  <alternate: [#]?> 
  <padding: [0? num]?> <precision: [.num]?> <modifier: [h|hh|l|ll|L|z|t|j]?>
  <format: [c|d(i)|e(E)|f|o|p|x(X)|u|s|g(G)]>

Is the order and meanings correct in the above? A couple examples being:

printf(" %-10.3s %-+20ld", "Hello!", 14L);

Answer 1

Is the following a correct understanding of the possible printf formats?

"Generally" yes, but for example you "can't" do %jg or like %0#p .

There is also %n .

Both "precision" and "padding" may be asterisks, like %*s or %.*s (but you could have defined num as ([0-9]+|\*) ...).

Also . is optionally followed by a number. So it's more like <precision: [. num? ]> <precision: [. num? ]> <precision: [. num? ]> - if only . is specified, precision is taken as zero.

Is the order

The order of - +#0 is irrelevant and you can repeat them, so you can %-+020d and %+0-+++000----20d with same meaning (and 0 is ignored when used with - , so also there are corner cases).

meanings correct in the above?

There is no explanation in the above. - is not "justification" (taken literally, a word?), it's a flag that makes the output be left justified within the field . Also meaning depends on context - "precision" for floats maybe can be understood as the number of digits after comma, but "precision of a string" sounds strange. But generally, yes.

Answer 2

Your specification is too restrictive:

• the flags + , - , # , 0 and space can appear in any order, but some combinations are meaning less, such as %+s .
• width and precision can be specified as *
• a and A were introduced to produce hexadecimal floating point representations
• F is available and different from f for NaNs and infinities.
• %% and %n should be recognised too.

Here is a regular expression to match all valid printf conversion specifications, but that will not detect invalid combinations:

%[ +-#0]*{[*]|[1-9][0-9]*}?(.{[*]|[0-9]*}?)?{h|hh|l|ll|L|z|t|j}?[%naAcdieEfFopxXusgG]

You might refine it to reject any flags for %% and restrict other cases too, but it will become quite complicated to express as a regex.