[mathiasrw]

Closes #856

Stacked on #860. In place of #857, after the review of 26 October.

Change

• rr has the values of RexReplace with long names, directly on it:
  • the file: filePath, filePathRel, dirPath, dirPathRel, dirName, fileName, name, ext, bytes, size, mtime, ctime, mtimeObj, ctimeObj
  • the numbers: fileNumber ($n), filesTotal ($t), matchedFileNumber ($i), changedFileNumber ($c), matchThisFile ($m), matchUntilNow ($M)
  • text, find, stdin (the piped data), cwd, time, timeObj
  • the groups: rr[0], rr[1] and rr.year, which overwrite a name they share
• print() and printErr(), also as rr.print() and rr.printErr(), print to stdout and stderr with a line break, also with --quiet-total. print() without rr runs once per file, so it can print once for a file.
• pad(value, width = 3, char = '0'), like pad($1) for 007.
• file_path and file_path_rel, the same as file and file_rel, with their space cousins.
• Code that names rr runs once per match, as it can reach a group in a way the search of the code can not see, like rr[name]. rr is found as a whole word, so {...rr} and const r = rr count too, and brr does not.

rr is a plain object: every value is its own, so {...rr}, Object.keys(rr) and JSON.stringify(rr) have them all. It is only made when the code names rr, and the values of the file are made once per file. 20 000 matches: $1.toUpperCase() takes 205.5 ms, and rr[1].toUpperCase() 205.6 ms.

Tests

test/fix/856.test.js is new, and 15 of its 17 tests fail without the change. It covers the groups by number and by name, the long names of a file and of the numbers over three files, text, find, cwd, time, stdin, a group that overwrites rr.name, a spread of rr, the run per match, that brr.n is not rr, print, rr.print, printErr and rr.printErr with --quiet-total, pad(), and file_path. The help and the README tell all of it. yarn test-ci passes.

From the code review of #855

The README row of --js says that under the flag n only the code has no $1, while a replacement text still reads $1 as the first named group. It is here, as this pull request changes that row too.