sqlite3: fix parser

[milahu]

continue #640

lazy pages based on #640 (comment)

rename fields from *_index to idx_* vaguely based on style guide via #640 (comment)
these are not physical offsets, so im not using ofs_*

migration script:

#! /bin/sh
# \b = word boundary
sed -i -E '
  s/\block_byte_page_index\b/idx_lock_byte_page/;
  s/\bfirst_ptrmap_page_index\b/idx_first_ptrmap_page/;
  s/\blast_ptrmap_page_index\b/idx_last_ptrmap_page/;
  s/\bptrmap_max_num_entries\b/num_ptrmap_entries_max/;
  s/\bvariable_size\b/len_blob_string/;
' "$@"

will squash commits later

[edzillion]

I tried to get kaitai struct working in lua and ended up here. It seems PR640 started work on fixing some of the problems but it was never merged. In any case I was able to parse my example db to some extent. It should be mentioned more clearly that since the pages are lazy loaded you will never see any actual pages, just the database_header. It took the kind people over in the kaitai struct gitter lobby to clue me in on that one. The lua documentation is very deficient. I note there is mention of this elsewhere but there is so much information to consume that having some kind of basic language specific tutorial on how to get the table data from your db would a much more direct way to get to grips with kaitai.

In any case, I am having issues with overflow pages. My example DB has page size of 4096 and page 2 is of type table-leaf, there is only one cell in the page and it says in the payload of this cell that the payload size is 7756. It says the overflow page number is 6. When my sqlite3.lua file processes this page it does not add a pointer from page 2 to page 6 like I am expecting, and if I look at page 6 it has 99 cells and all they is one property each: ofs_content:27769

how to I access this offset content? I stepped through the parser and I my crude understanding is that there should be a Sqlite3.BtreePagePointer() created to link this page with it's overflow page on page 6.

When I look at page 6 it has 99 cells and each contains one property ofs_content:27769 which I guess is a memory offset?

What's the state of this PR, can I do anything to help it along?

edit: in fact, I would be happy to contribute a Lua tutorial, if I can get it working.

[milahu]

What's the state of this PR

abandoned

can I do anything to help it along?

add example code (python, lua, ...)

i needed this sqlite parser for my pysqlite3 to parse a partially-downloaded sqlite database
the full database has 130GB, so for my case, the pages must be lazy

kaitai does not support such lazy parsers
so using this parser requires some manual parsing

kaitai is just a code-generator
and sometimes the generated code needs patching

I am having issues with overflow pages.

overflow means, data is stored on multiple pages

see connection._table_values

payload is overflow_record
must be parsed manually because paging

cell.ofs_content is also used in connection._row_locations
to get the raw byte offsets of values
i used this to build an external index, to map row ids to byte offsets
the database is shared over bittorrent
and with the index i can map row ids to torrent pieces for partial download

[edzillion]

Yeah, I am trying to parse 'app size' sqlite dbs on limited hardware so lazy parsing isn't an issue. That's an interesting application you have, thanks for the pointers, I will have a look.