Speed up hts_parse_decimal() handling of oversized exponents

The loop to apply exponent values could be made to run for a
very long time (although not forever) if the exponent value
in the number passed to the function was very large.  As in
both the +ve and -ve exponent cases this will eventually
result in `n` becoming zero, it's possible to short-cut the
rest of the loop as the final value will no longer change.
(While the -ve case is obvious, the +ve one is a bit more
complicated due to wrap-around.  However, as each
multiplication by 10 adds a factor of 2 (as 10 = 5 + 2),
after at most 64 enough 2's will have accumulated in the
product to make it divide exactly into 2^64 and the result
will be zero.)

Note that this does not try to add overflow detection, which
would require more extensive changes and is left to future
work.

Signed-off-by: Rob Davies <rmd+git@sanger.ac.uk>

Author: daviesrob

NEWS

@@ -94,6 +94,9 @@ Bug fixes
   BGZF::errcode field after calling bgzf_close() as it may no longer be valid.
   (PR #2035, #2042)
 
+* Speed up hts_parse_decimal() handling of oversized exponents
+  (PR #2045)
+
 Documentation updates
 ---------------------
 

hts.c

@@ -3923,8 +3923,8 @@ long long hts_parse_decimal(const char *str, char **strend, int flags)
     }
 
     e -= decimals;
-    while (e > 0) n *= 10, e--;
-    while (e < 0) lost += n % 10, n /= 10, e++;
+    while (e > 0 && n) n *= 10, e--;
+    while (e < 0 && n) lost += n % 10, n /= 10, e++;
 
     if (lost > 0) {
         hts_log_warning("Discarding fractional part of %.*s", (int)(s - str), str);