Org Merge Driver To-do List

Webpage

Add the original posting to the proposal page

State “DONE” from “NEXT” [2012-05-09 Wed 01:36]

Fix all references to the project to be the same name on the webpage

Fix the reference to the git repository after it has been created

Update the usage instructions

Add implementation notes

even though nothing is decided, there must be something I can add

Prototype

fix up documentation

State “DONE” from “TODO” [2012-05-09 Wed 01:34]

Write real manual

State “DONE” from “NEXT” [2012-05-09 Wed 01:34]

manual, how to use it
now in the root README

Finish the parser

State “DONE” from “NEXT” [2012-05-10 Thu 19:03]

make sure the new headings are being put in the right spot
- means: compare level of new heading with old heading to add it in the right spot
- possibly add the new heading to the propper spot right as the next one is started

Implement a heading tree printer

State “DONE” from “NEXT” [2012-05-09 Wed 23:24]

really a org_document_to_file function

Add a TAP for the program which demonstrates its use

test anything protocol

Write an introductory warning

State “DONE” from “NEXT” [2012-05-10 Thu 19:03]

this program assumes things about the structure and use of org-mode and org files. If the way you use org files does not match this, then it may corrupt your data! Please be very aware of what assumptions this program makes before using. I (nor anyone other than yourself) can be held accountable for the results of using this program.

Use Gnulib diffseq to merge

State “DONE” from “NEXT” [2012-05-16 Wed 22:26]

Fix memory usage

nothing is being unmalloced
write destroy functions
- one for document, heading, look at ?gl_list_free? for list notes

Start working on the change detection

State “DONE” from “NEXT” [2012-05-16 Wed 22:25]

Create a heading mapping

Begin work on the change merging

Create text mapping

Allow headings to be ‘moved’ locally

Allow headings to be ‘moved’ globally

Print propper merges

Finish the prototype

Org Merge Driver

[#B] Argp on main

[#A] Create doc_stream sdio wrapper

write a stdio wrapper
make sure that we can encode

[#B] Make the parser construct the doc trees

[#B] Merge tree

implement the prototypes
??

[#C] Memory allocator for document

[#A] Support global elt matching

Allow elements to deal with global matching

Add callbacks so elements can choose if and how they match.

[#A] Update autotool files

[#C] Add support for multiple encodings

[ ]
[ ]

Devise a doc_elt data allocation scheme

Currently:
- There is no standard scheme for setting up data in an element
- org_heading accepts a pointer to an outside allocated buffer, while org_text actually copies and appends strings passed to it.
Ideas:
- Outside allocation, pass buffer to const data.
  - who deallocates?

Fix printmerge context

Fix org_lexer context

Currently using a mishmash of global vars in a reentrant parser.
Make a lexer userdata struct to use

[#B] Make doc_stream macros safe

[#B] Check and handle memory allocation errors

Currently, datastructures malloc without returning error.
not detected, checked, etc.
See how gl_list does it
- Most of our interfaces are modelled slightly afer gl_list

[#B] Make doc_merge work with null ltree_lists

merge_node and doc_nodes currently always have allocated child lists.
They should only have one if they have children

[#A] Implement org_heading

[#A] Fix the merge_delta/doc_merge interface

Currently uses status??
Make a real enum & field, don’t overload something else.

[#B] Fix Parser / Lexer

currently:
- parser is implemented inside the lexer
- lexer logic is more complicated than it should be
- parser is almost empty
To fix:
- Make lexer return smaller symbols
- Build doc_elt’s inside the parser
Why bother:
- Adding rules to the lexer is error prone & hard.
- May make the parser more flexible.

[#B] Move phony_elt to it’s own header

phony_elt is a dummy elt used for testing.
moving it to it’s own header file will let us use it in multiple tests.

[#A] Create a debug print macro

support different modules enable/disable
- eg disable printing in document elements, enable in test functions.
fail message macro?
- disapears like assert?

[#C] Create a verbosity-controlled print function

[#A] Make doc_elt_compare find updates

doc_elt_compare returns a simple boolean
make doc_elt_compare decide if something matches
- if they do, let it decide whats newest / if something has been updated
If compare has to compare every field of a doc_elt, it only makes sense to cache that data for printing.

Code cleanup

Ideas

(andrew) add and implement functions to create the mapping and dnode tree

implement org_text.h,r org_document.h, org_phony.h

make the parser work for those elements

Long Run

Log

[2012-05-07 Mon]

modify gl_list.h to not inline functions
write parser, reads the files but does not create a proper list

[2012-05-08 Tues]

split source files

the files should really be split up by their use

split commits

branch for the prototype
add documentation
add gnu_lib list
add read file
add main program
- with only org_ structs
add parser

finally started uploading code

[2012-05-09 Wed]

tried to add some testing code
- went with TAP at first, but decided it wasn’t what I wanted
- went with heavier duty unit testing (called ‘check’), but it was too heavy duty
- trying to go back to TAP
tried to compile the parser into a library instead of statically linking into the program
- got it working but reverted; this is more useful for the final project (and with the unit testing removed there was no point)
wrote a manual for using the program (for when it is eventually done)
tried to make all the random bits of (evidently pointless) work presentable

[2012-05-10 Thu]

looked into anonymous functions in C, definitely not portable
wrote a function to recursively call other functions on a tree
wrote a function to print a org_doc back into a file
finally finished the parser, i probably made it more correct than necessary to show how it would work
fixed a lot of the documentation in the code I have written, adding GPL warnings at the top of all my source files
finally fixed the problem with gl_list not compiling inline functions. I needed to include autoconf’s config.h before it to define the HAVE_INLINE macro.
starting to think about the best way to finish off the prototype
- this involves the ‘tricky’ part of the program: finding changes and merging them.
looking into different parser generators for the final program. I am worried about both speed and robustness when it comes to creating a parser

[2012-05-11]

looking into more robust parser generators, specifically ones that will be extensible, easy to use, and most importantly fast
typically Bison and yacc look very good

[2012-05-12]

reading about compilers, trying to see if the traditional structure will lead to hints as to how to set up my code
reading a book known as ‘the dragon book’
found out that you can collapse a heading which is inside a #+begin_src: block
- is this a bug?

[2012-05-13]

spending some time going through the manual again to figure out what

[2012-05-14]

thinking about the best way to set up a difference detector.
I am finding that a lot of ideas I have about what rules to use to merge might not be desirable for people
how can I set it up, so that the rules can be customized?

[2012-05-15]

After reading papers about ‘sequence difference seeking algorithms’, I decided that it would be too much work to implement on from scratch for the prototype
Found an implementation in Gnulib (should have checked there sooner!)
spending my day attempting to include it into the project

[2012-05-16 Wed]

deciding that for the prototype, I will only do change detection on a heading level
- this makes for kind of a bad prototype
- I will have to really make sure I consider how having a larger amount of elements will affect the project
it is really confusing trying to match headings when one is not unique
it will be necessary to create a mapping of headings in from one file to another, where the mapping says which headings are the same, which are non-unique non-distinguishable, and which have no matches

[2012-05-19 Sat]

starting the change detection
must finish the fucntion to create an empty tree of mappings from the origin file
- this will involve copying the code from thre recursive function and manipulating it to create the treec