Consolidate number parsing

[mlinkert]

Right now, number parsing in general is not very consistent. In some places, we have something like this:

  Double foo = new Double(bar);

...with no error checking. In other places, we do this:

  try {
    Double foo = new Double(bar);
  }
  catch (NumberFormatException e) { }

In still other places:

  try {
    Double foo = new Double(bar);
  }
  catch (NumberFormatException e) {
    LOGGER.debug("Could not parse double value", e);
  }

And finally, in some places we attempt to sanitize the String being parsed (to account for differences in the decimal separator):

  Double foo = new Double(DataTools.sanitizeDouble(bar));

It would be nice to standardize how this is done, so that a combination of the last two approaches is used across the board. This means adding methods to loci.common.DataTools?:

  Byte parseByte(String s);
  Short parseShort(String s);
  Integer parseInteger(String s);
  Long parseLong(String s);
  Float parseFloat(String s);
  Double parseDouble(String s);

...which encapsulate all of the parsing, error checking, and logging and will return null if the String could not be parsed.

The readers then just need to be updated to call the appropriate method in DataTools? instead of constructing number objects directly.

[jmoore]

See also the discussion under ​https://github.com/openmicroscopy/bioformats/pull/50 for parsing attributes, but that would mean that these files might need to also use BaseHandler:

git grep "import org.xml.sax.helpers.DefaultHandler" | wc
       9      18     912

[mtbcarroll]

The proposed utility methods should use the standard JDK static methods rather than new as in many cases a new allocation can be avoided. (Also, .trim() is often not needed as that is handled automatically.)

[mlinkert]

Might be worth considering splitting this into two parts: the new API in DataTools? could be added for 5.2.0 (this should be a fairly quick task), and readers could be updated as time permits in 5.2.x+. The new API methods probably shouldn't be added in a patch release, so that would let this move forward without having to wait for 5.3.0.

[jamoore]

Leaving it to Seb whether the new API fits on the 5.2.0 board.

[sbesson]

See ​https://github.com/openmicroscopy/bioformats/pull/2285 for the API addition

[sbesson]

See ​https://www.openmicroscopy.org/qa2/qa/feedback/17150/ for a reader where this new API could be implemented.

[jamoore]

Will the QA be lost if this is closed?