jwutils.log.set_flags() and .set_level() is not set until after all commands are loaded. This can make debugging complicated. OTOH, command-line parsing is also not done until all commands are loaded, because they define the command line options. This commit adds some poor man's command line parsing for fetching --log-level and --log-flags only, and applying them before the commands are all loaded.
Add support for adding children of type StringTree. If a StringTree argument is passed to .add(), this should do the right thing. It makes use of the newly added .__add_children() method and involves quite a bit of hairy case distinctions.
parse() and read() up to present automatically set the root content of a StringTree object to "root". This can now be optionally defined by the "root_content" keyword argument.
- Add functions: append_to_prefix(), remove_from_prefix() set_filename_length() - Make slog_m() log one list item per line - Add console_color_chars(prio)
This commit makee Cmds run all sub-commands in an asyncio event loop. The event loop is currently passed to the commands, which seems unnecessary and looks like it's using a feature which is bound to be deprecated in the future.
object_builtin_name() is meant to derive an object name from its class name. Doesn't work well, but adding the code nonetheless for improving it later.
This commit flips some more switches from Python 2 to Python 3 in makefiles and Python code. Build runs through, but it's still likely to break things.
syntax, i.e. non-assignments, without equal sign. Half-baked, but still better than before. Also support some more Python 3, i.e. dict.items() instead of iteritems()
o basestring is merged into str o print() needs parentesis o Class inheritance syntax changed o Abstract baseclass (ABCMeta) syntax changed o map.iteritems() is replaced by map.items() o Inconsistent use of tabs and spaces are no longer tolerated
run_sub_commands() was the only way to access subcommands up to now, Cmds.run() adds another interface, an object as a place to add customizations affecting all commands, e.g. global command line options.
This replaces the mechanism to always construct lists with a leading empty rule. This leads to many RR-conflicts, so this empty rule is omitted and later added conditionally by grammar_fix_list_recursion(). Sadly, the condition is bollocks, since it's always true.
First time parsing doesn't error out with a syntax error. No usable AST is produced, strings are not returned from lexer, and AST lists aren't lists, really.
TEXT:="Hello world!"; had to be excluded from the example, because I don't get how this could be parsed with the given syntax. There's a special sequence "all visible characters", but any lexer regex I could think of will also match the types defining "alphabetic character" and return the respective tokens (e.g. T_A) or vice-versa, depending on the order in the lexer input file. I suppose, the only sensible way to handle this, is to define "all visible characters" by defining the tokens for the missing characters, and then use them along T_A ... T_Z or their derived types.
grammar.py is meant as a compiler compiler compiler, taking in a grammar structure and returning flex and bison files. It can also parse EBNF. Currently there are still remnants of VHDL-specific stuff in the code, because parsing VHDL was what it was initially about.
