flat-text MARC editor doesn't handle null indicator values

[Galen Charlton]

Evergreen version: trunk@19040 (assumed to apply to rel_2_0 and rel_1_6 as well)
OpenSRF: 1.6
Postgres: 8.4
Linux: Debian Squeeze

When creating a MARC record from a template using the flat-text editor, null indicator values are not handled well. For example, using the default k_book template, the initial value is:

=LDR 00620cam a2200205Ka 4500
=008 070101s\\\\\\\\\\\\\\\\\\\\\\\\\\\\eng\d
=010 $a
=020 $a
=082 04$a
=092 $a
=100 $a
[etc.]

This lacks the indicators on the variable data fields. If you don't notice this and input data in the record, e.g.,

=LDR 00620cam a2200205Ka 4500
=008 070101s\\\\\\\\\\\\\\\\\\\\\\\\\\\\eng\d
=100 $aSimpson, Homer
=245 $aDoh! An autobiography.

the resulting record looks like this after saving it and re-opening it in the flat-text editor:

=LDR 00620cam a2200205Ka 4500
=001 5009
=003 CONS
=005 20101222163307.0
=008 070101s\\\\\\\\\\\\\\\\\\\\\\\\\\\\eng\d
=100 \a$aSimpson, Homer
=245 \a$aDoh! An autobiography.
=901 \\$a5009$b$c5009$tbiblio

Note that the second indicator in the 100 and 245 is (incorrectly) set to $a.

Similarly, if you edit a record in the original MARC editor, add a tag, tab through the indicators instead of entering spaces, the resulting tag looks like this in the flat-text editor:

=246 $aDoom

Expected behavior:

[1] When a record is opened in the flat-text editor or seeded from a template, indicator values that are the empty string should be converted to spaces.

[2] Arguably, if the user forgets to enter indicators in the flat-text editor for a variable data field, they should be initialized as blanks. E.g., the editor can look for the pattern /^=(\d{3}) \$/ and (for tag values >= 010), automatically insert the two blank indicator values.

[Mike Rylander (mrylander)]

The solution proposed in [2] is complicated by the fact that the subfield separator is configurable, but by checking to see if breaker input has a subfield separator in position 4 we can insert the needed indicator placeholders.

This means that users can do either of the following independently:
  * forget to supply indicators
  * fail to supply a leading subfield delimiter

but not both together -- we can't tell the difference between those states programatically.

Implemented in r19049.

[Mike Rylander (mrylander)]

Note, until I have confirmation of the efficacy of r19049 I won't be backporting to a release-generating branch such as rel_2_0 or rel_1_6_2. Testers, please!

[Galen Charlton (gmc)]

I've tested Mike's patch; added a follow-up of my own, and backported to rel_1_6, rel_1_6_2, and rel_2_0.