Kevin Day [Sat, 9 May 2020 22:26:41 +0000 (17:26 -0500)]
Feature: add f_string_static as a non-allocated alternative to f_string_dynamic
That is not to say that a dynamically allocated buffer cannot be placed in its string part.
Instead, it communicates that this is not never be allocated or deallocated directly.
This is implemented in such a way that f_string_dynamic is a sub-type of f_string_static.
This should allow for easy type conversions.
Kevin Day [Sat, 9 May 2020 21:47:12 +0000 (16:47 -0500)]
Update: improve color mode handling
Improve handling of color mode so that applications can consistently react to the established color code.
In the fake program, this is useful for passing the color mode to the build pre-process and build post-process scripts.
Kevin Day [Sat, 9 May 2020 04:28:31 +0000 (23:28 -0500)]
Update: add snatch apart functions and improve logic of fss snatch functions
The snatch apart is to create separate sets of strings instead of merging them.
The snatch mash apart will mash all dynamic strings for all content for a single object but use separate strings for each separate object.
I accidentally, or perhaps habitually, used break when return is better when handling errors.
Make sure to document f_string_max_size as a possible return result.
Kevin Day [Sat, 9 May 2020 01:23:46 +0000 (20:23 -0500)]
Update: switch back to lengths from inclusive ranges
Using start and stop ranges ended up being cumbersome.
Using length is simpler and a start range can be used by just adding to the source string before calling.
Kevin Day [Fri, 8 May 2020 02:47:18 +0000 (21:47 -0500)]
Update: numerous fixes and additions, mostly string related
Not sure how I let this one get through, but add missing return statement in a couple of private string functions where it is missing.
In retrospect, it seems that I was overzealous with some of the parameter checks.
Redo the parameter checks to not be as sting on string length and buffer size ranges.
Instead, just return that there is nothing to do instead of erroring out.
Add string terminate functions.
These functions are designed to guarantee that an end of string exists at the end of some dynamic string.
Add functions for appending one or more arguments to the arguments array for the execute functions.
This functionality is common enough (used extensively by firewall and soon to be used by fake).
Support providing an array of arguments to append.
This greatly simplifies long-winded code.
Minor fixes and cleanups in the execute functions:
- Add missing parameter checks.
- Rename results to result to be more accurate.
Kevin Day [Wed, 6 May 2020 04:14:13 +0000 (23:14 -0500)]
Progress: begin re-implementing fake program (Featureless Make)
The goal here will be to mimic the current FSS build scripts fist.
Then update the FSS build scripts to the more generalized design.
Then implement the traditional fakefile process (a makefile alternative).
Currently has "skeleton" mode, which may or may not be implemented.
This was added as a stub just in case I decide to implement it.
Kevin Day [Tue, 5 May 2020 05:00:50 +0000 (00:00 -0500)]
Feature: add fss snatch functions for "snatching" content for a set of object names
These are intended to be used for simple purposes.
Directly implementing these in individual projects are recommended for more advanced usage.
I intend to expand on these "snatch" functions in the near future.
Kevin Day [Tue, 5 May 2020 03:01:27 +0000 (22:01 -0500)]
Update: add back the dynamic_partial functions, add missing functions
After further review I realized that the dynamic partial behavior is going to be more common.
The rip functions are also not the same because it removes leading/trailing whitespace, which may be undesirable.
Some of the append_nulless function implementations were missing.
Kevin Day [Mon, 4 May 2020 11:47:38 +0000 (06:47 -0500)]
Feature: add fl_utf_file project
This is intended to be an f_utf_character handling version of fl_file project.
This f_utf_character is a 4-byte wide storage for a single UTF-8 character.
I accidentally committed the utf_file build settings in a previous commit: f69e57a.
Kevin Day [Mon, 4 May 2020 11:11:59 +0000 (06:11 -0500)]
Update: Redesign f_file and fl_file functions, simplifying f_file_position
Redesign the logic behind the f_file and fl_file functions to favor the buffer->used append strategy.
Remove f_file_read_range(), something similar can be added later if need be.
Povide f_file_reat_until(), that functions similar to f_file_read_at(), except the file position is not automatically set.
Using f_file_reat_until() is more efficient in array loops because it does not check the file position every execution.
Add f_macro_file_seek_data() and f_macro_file_seek_hole() to support the newer seek modes: SEEK_DATA and SEEK_HOLE.
The f_file_position structure is simplified.
The file position should only be for file related code and having buffer in there is just plain wrong.
The structure is further simplified in its property names.
Kevin Day [Mon, 4 May 2020 02:43:46 +0000 (21:43 -0500)]
Update: make fl_file functions more consistent and update documentation
Refactor the fl_file functions.
Rewrite memory management in f_file and fl_file functions.
Fix bug with size of bytes, renaming it to byte_chunk to avoid confusion.
Add block size property to f_file structure to allow custom block size handling (as opposed to the previous behavior of using a hard-coded default).
Kevin Day [Sat, 2 May 2020 23:42:44 +0000 (18:42 -0500)]
Update: simplify and rework string functions
The rip and append functions are functionally identical.
Remove the rip function.
With the rip functions gone, the rip_trim functions are renamed to rip.
This way "rip" functions operate like a theoretical append_trim function, but with fewer words.
Make the string functions use start and stop ranges instead of lengths.
This eliminates the needs for the partial functions.
Add new nulless functions that check for and remove nulls from source strings while transferring them to the destination string.
This allows for converting the strings that may have nulls to nulless strings such that they can be used in null-sensitive functions such as those found in the standard C libraries.
Make sure the utf_string functions match correspond to the string functions.
Kevin Day [Sat, 2 May 2020 16:07:19 +0000 (11:07 -0500)]
Update: at f_array_too_large status and don't include terminating NULL in string lengths
Just like f_string_too_large, there needs to be a generic f_array_too_large to report that the array is larger than can be supported.
Also provide an f_array_max_size #define just like f_string_max_size.
The #define string lengths shouldn't include the terminating NULL.
This will prevent having to remove the NULL when converting to/from null-agnostic to null-aware designs.
Kevin Day [Sat, 2 May 2020 05:10:23 +0000 (00:10 -0500)]
Feature: add word, dash, and plus UTF-8 character checker
The use of the '+' operator is more common than in the past.
Add support for detecting if a character is a word character, a dash character, or a plus character.
Kevin Day [Sat, 2 May 2020 04:51:35 +0000 (23:51 -0500)]
Update: Fix string length calculation and report when no changes for fll_program functions
Both fl_string_rip and fl_string_rip_trim use include ranges and not lengths.
Record the start length to identify if there were changes and return f_no_data when there are not changes.
Kevin Day [Fri, 1 May 2020 04:56:26 +0000 (23:56 -0500)]
Cleanup: restructure bash script cleanup process
Explicitly return 0 in main.
The cleanup function can be called within the functions being removed.
This is handy in that the cleanup no longer has to be called after the main function.
The error code can then be properly returned without being lost due to the cleanup function.
Once this is done, return can be used instead of exit (which is generally safer).
Kevin Day [Fri, 1 May 2020 03:36:45 +0000 (22:36 -0500)]
Update: implement missing serialized functions and fix bugs
Implement fl_unserialize_simple() and fl_unserialize_simple_get().
The last character in the last string in the unserialized array gets cut off.
- This happened because the stopping point was incorrectly compensating for the splitter character for the last character in the string.
The width (UTF-8 character width) is not part of the locations size.
Kevin Day [Tue, 28 Apr 2020 05:46:11 +0000 (00:46 -0500)]
Feature: implement string mash support
String mashing is a way to append a string to another with a glue string in between.
The idea is that a space could be placed between the two strings.
A string is primarily used as the mash character so that UTF-8 can be natively supported as the glue character.
Kevin Day [Mon, 27 Apr 2020 02:13:28 +0000 (21:13 -0500)]
Update: short variable name, reorganize function, fix types
The use of parameter_counter ended up being very wordy with "parameters.parameter[parameter_counter]".
- Change this to 'i', since that is unused and traditional.
Reorganize the additional parameter processing loop.
Use uint8_t instead of "unsigned short" for more consistency.
The width_max should be f_number_unsigned and not "unsigned short".
Add additional documentation for console ids in regards to what "empty" means.
Kevin Day [Sun, 26 Apr 2020 07:14:39 +0000 (02:14 -0500)]
Bugfix: remove null pointer check from file stat
Do not assume that stat was passed as an uninitialized pointer.
In the case of passing a class by reference, the stat pointer would be non-zero.
This pointer is not dynamically allocated and therefore not an error.
The null pointer check is therefore invalid for these cases.
Kevin Day [Sat, 25 Apr 2020 17:46:36 +0000 (12:46 -0500)]
Update: redesign byte_dump --last to be inclusive
The --last parameter should be inclusive for consistency with the rest of the project, namely FSS.
By redesigning the --last to internally be represented as a relative offset, the behavior can be simplified.
Doing this also adds support for potential future designs where a --length parameter may be provided that is a relative size parameter instead of an absolute one like --last.
Kevin Day [Sat, 25 Apr 2020 01:54:05 +0000 (20:54 -0500)]
Update: add initial handling of zero-width space in FSS projects
UTF-8 zero-width characters have the potential for being combining characters.
In such cases, the combiner/joiner should be considered part of what is being combined.
That is a combiner/joiner before a graph should be treated as a graph.
A combiner/joiner before a whitespace should be treated as a whitespace.
Disclaimer: I suspect that this will eventually need to be broken down to handle each specific case.
A combiner/joiner on whitespace that results in rendering a printable/visible character would be a violation of the whitespace design principles of FSS.
Further investigation is needed and will likely require changes.
Add appropriate @todo for further development of this functionality.
Rename max_width to width_max to follow newer practices.
Kevin Day [Sat, 25 Apr 2020 00:39:35 +0000 (19:39 -0500)]
Feature: add f_string_length_size max of (uint64_t - 4)
The goal here is that any string processing must be able to add a given UTF-8 width (4-byte) and not oveflow.
Much of the code in this project will not check this as it should be done so at a higher level (performance reasons).
The ideal time is that when allocating some string, always allocate at max f_string_length_size.
Kevin Day [Fri, 24 Apr 2020 02:43:01 +0000 (21:43 -0500)]
Bugfix: fix UTF-8 whitespace detection and provide zero-width detection function
The whitespace detection codes for UTF-8 were incorrect.
Non-printing characters, called zero-width, are not whitespace.
Move them out of the whitespace detection and provide a new function for detecting zero-width.
Handle additional UTF-8 whitespace character codes that I had previously missed.
Kevin Day [Thu, 23 Apr 2020 01:40:50 +0000 (20:40 -0500)]
Feature: implement support for the -T/--trim parameter
Provide support for trimming the object names on input and output.
After implementing this I suddenly remember that the standard might require that the whitespace before and after a valid object name are to be ignored.
This may be removed in the future and fixed in the library.
Additional investigation on how I want to handle this needs to happen first.
The standard is originally designed around ASCII, which only ASCII whitespace is considered whitespace.
This will probably have to be fixed to match the additional goals of the project in terms of whitespace handling.
Kevin Day [Wed, 22 Apr 2020 02:00:18 +0000 (21:00 -0500)]
Feature: expand string functions and utf-8 string functions
The need for this was realized while developing this trim parameter feature.
This has been added as is for commit isolation and may be incomplete.
This utilizes private functions to reduce duplicate code.
While the use of private functions is generally unwanted in this project, this specific case seems to be an exception.
Add rip functions.
Add non-dynamic equivalent of some string functions.
Add trim functions.
Kevin Day [Tue, 25 Feb 2020 03:28:38 +0000 (21:28 -0600)]
Progress: continue development of FSS Extended List
In particular:
- Remove excessive fl_fss_increment_buffer() uses.
- The removed code may be a good idea long term, but for now use a simpler and more efficient approach.
- Fix some mistakes in the slash delimiter handling.
- Begin the initial work for recursively (or so..) processing the nested lists.
Kevin Day [Sat, 23 Nov 2019 04:56:20 +0000 (22:56 -0600)]
Update: build level and monolithic fixes and improvements
Add missing library: fll_file.
Make it even easier to compile against "level" and "monolithic" build processes by providing "--level" and "--monolithic" parameters to the generate.sh script and associated settings files.
Make sure package.sh clears other build modes (when --level is specified, make sur --individual and --monolithic modes are not set).
Kevin Day [Fri, 22 Nov 2019 02:24:43 +0000 (20:24 -0600)]
Update: implement *_delete_simple() and *_destroy_simple() macros and related changes
I do not want a *_delete() and *_destroy() that doesn't provide the ability to handle a status code.
However, in practice, the *_delete() and *_destroy() macros rarely needed the status response checked.
Additional f_status variables were created only to be provided for the status parameter of the stated macros.
This is a waste.
This provides and utilizes alternative *_delete() and *_destroy() macros called *_delete_simple() and *_destroy_simple().
These simple macros do not accept or process the status code.
The resulting code is simpler and easier.
I am on the fence whether or not to throw away the *_delete() and *_destroy() macros that utilize a status parameter.
Future versions may or may not replace the *_delete() and *_destroy() with the *_delete_simple() and *_destroy_simple() macros.
Update *_delete() and *_destroy() macros as needed.
Kevin Day [Wed, 20 Nov 2019 01:43:54 +0000 (19:43 -0600)]
Refactor: make status codes more consistent
Better follow the naming paradigm where reasonably possible.
Use more consistent naming, which should also help with reducing the changes of enum to function name conflicts.
Alphabetically organize status codes per group.
Kevin Day [Wed, 20 Nov 2019 01:00:00 +0000 (19:00 -0600)]
Update: rewrite status code functions and related changes
Fix several problems with the status code processing, namely missing or incomplete data.
Add status code max size defines.
Add missing "#ifndef _di_f_status_codes_" to status codes.
Update cases where 'error' was not renamed to 'code' in status code sources.
Add missing status codes.
Fix incorrect define, "#ifdef _di_fl_status_invalid_" should instead be "#ifndef _di_fl_status_invalid_".
Fix incorrect sizes associated with status code output strings.
Prepare pipe support (not yet implemented).
Ensure f_utf is included and linked in all appriopriate dependencies.
Treat f_utf as a core level_0 project and update documentation accodingly.
Relocate fl_console_parameter_process() into f_console_parameter_process().