Nice work! Mostly nit comments and had 2 questions

Reviewed 2 of 14 files at r3, 38 of 38 files at r5, 38 of 38 files at r6, all commit messages.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @otan, @renatolabs, and @srosenberg)

pkg/sql/pgrepl/pgreplparser/lexer.go line 84 at r6 (raw file):

	case ch == doubleQuote:
		l.scanUntilEndQuote(lval, doubleQuote, lexbase.NormalizeString, IDENT)
	case unicode.IsDigit(ch):

nit: This case overlaps a lot with the isID(ch, true) case for the LSN check. I wonder if they can share that part of code.

pkg/sql/pgrepl/pgreplparser/lexer.go line 91 at r6 (raw file):

			switch {
			case unicode.IsDigit(nextCh):
			case isHexDigit(nextCh) || nextCh == '/':

if unicode.IsDigit(nextCh) == true, is isHexDigit(nextCh) always true? If so, shouldn't we remove the case unicode.IsDigit(nextCh)? Otherwise, these digits will never be parsed as an LSN.

pkg/sql/pgrepl/pgreplparser/lexer.go line 150 at r6 (raw file):

}

func isID(ch rune, firstCh bool) bool {

I had some trouble understanding what ID means here. Could you add some comments to this func, and maybe have 200 and 377 as const var?

pkg/sql/pgrepl/pgreplparser/lexer.go line 161 at r6 (raw file):

}

func (l *lexer) checkLSN(lval *pgreplSymType, startPos int) bool {

nit: it would be helpful to have comments explaining the expected format of LSN -- two hexadecimal numbers of up to 8 digits each, separated by a slash.

pkg/sql/pgrepl/pgreplparser/pgrepl.y line 1 at r6 (raw file):

%{

nit: for the whole yacc file formatting, tabs -> spaces

pkg/sql/pgrepl/pgreplparser/pgrepl.y line 218 at r6 (raw file):

/*
 * SHOW setting
 * NOTE(otan): we omit K_SHOW as this should be handled by the existing protocol.

Could you say more about this? Does it mean replication config params can only be queried via the pg protocol? Why?

pkg/sql/pgrepl/pgreplparser/pgrepl.y line 376 at r6 (raw file):

			K_TIMELINE_HISTORY UCONST
				{
					$$.val = &pgrepltree.TimelineHistory{Timeline: $2.numVal()}

nit: pg has a check for positive argument here. Do we want to have one as well?

pkg/sql/pgrepl/pgreplparser/pgrepl.y line 400 at r6 (raw file):

			K_TIMELINE UCONST
				{
					$$.val = $2.numVal();

nit: ditto, positive arg check

pkg/sql/pgrepl/pgreplparser/testdata/lexer/lsn.ddt line 39 at r6 (raw file):

bA101001010101010/f
----
ERROR: at or near "bA101001010101010/f": syntax error: error decoding LSN: strconv.ParseUint: parsing "bA101001010101010": value out of range

nit: maybe add a case that ends with '/'?

pkg/sql/pgrepl/pgreplparser/testdata/parser/start_replication.ddt line 57 at r6 (raw file):

----
START_REPLICATION SLOT slot_1 LOGICAL A0010/1011 (opt 'opt', "xXx" 'a') -- normalized!
START_REPLICATION SLOT slot_1 LOGICAL A0010/1011 (opt '_', "xXx" '_') -- literals removed

nit: maybe adding a case with an invalid LSN?

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @otan, @renatolabs, and @srosenberg)

pkg/sql/pgrepl/pgreplparser/pgrepl.y line 220 at r8 (raw file):

/*
 * SHOW setting
 * NOTE(otan): we omit K_SHOW as the fallback from being being parsed by

nit: remove one of the "being"

pkg/sql/pgrepl/pgreplparser/testdata/lexer/lsn.ddt line 37 at r8 (raw file):

# ends with a /
lex
00ff/

This doesn't seem to be supported by pg:

postgres=# SELECT '00ff/'::pg_LSN;
ERROR:  22P02: invalid input syntax for type pg_lsn: "00ff/"
LINE 1: SELECT '00ff/'::pg_LSN;
               ^
LOCATION:  pg_lsn_in, pg_lsn.c:74

postgres=# SELECT '00ff/00'::pg_LSN;
 pg_lsn 
--------
 FF/0
(1 row)

:

Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @otan, @renatolabs, and @srosenberg)