error message from ddc_index (tab-mode, multiple-files):
Indexing file `dump/import/../1.tabs'
the first line in tab-formatted file must be a document delimiter ("%$DDC:meta.file_=")
14:15:58.100450353 (ddc_index) > cannot index the project!
all tab-import lines beginning with "%%" which are not "special" commands ("%%$DDC:") should be ignored
even tab-import lines beginning with "%%$DDC:" which aren't "special" commands right now should be ignored (we might later want/need to introduce "%%$DDC:feature_foo=bar" and to compile that index with older binaries, e.g. during bug-hunting)
maybe just scan lines until you find a doc-delimiter and ignore everything until you find one?
bonus points: count the number of dropped lines and report it when you do find a doc-delimiter
in either case, blank lines, lines beginning with "%%", and lines consisting only of whitespace can and should be allowed (and ignored, and not counted) even before the first doc-delimiter in a file.
Bryan, in https://sourceforge.net/p/ddc-concordance/tickets/4/ I already done skipping empty lines and unknown %%$DDC:feature_foo=bar. But may be I do not understand you.
Right now %%$DDC:meta.file=somefilename.txt is a file border, so if you write
%%$DDC:meta.scan_=scan_bibl_stub
%%$DDC:meta.file_=some_file_name.txt
then scan_bibl_stub goes to the previous file. Is it a solution you want to change?
Last edit: Alexey Sokirko 2019-11-07
in single-file tab-input mode ("streaming mode"), then yes,
scan_=scan_bibl_stubshould go to the previous file in this case. in multi-file tab-input mode, it can either be ignored or assigned to the current file. The lines I'm actually seeing in the output ofddc_dump -f -tbefore%%$DDC:meta:file_=are:Ok, the lines before the first "%%$DDC:meta.n=" are always assigned to the first file. That must be true in the current source code.
Diff: