Skip to content

Latest commit

 

History

History
279 lines (198 loc) · 7.98 KB

zograscope.md

File metadata and controls

279 lines (198 loc) · 7.98 KB
Raw

NAME

zograscope is a collection of source-processing tools spawned out of syntax-aware diffing.

The nature of syntax-aware diff requires knowledge of structure of the code, which can be used to build other simple tools that can benefit from this information. Competing with real language front-ends in the level of accuracy is not possible, but making some things that are one step further than regular text-processing utilities seems feasible and the result might be even more practical than some of the more elaborate tools which end up requiring complicated setup process.

LANGUAGE SUPPORT

Language Status
C C11 and earlier with common extensions, but without K&R syntax
C++ C++14 and earlier with common extensions
Bash Not targeting a specific version
GNU Make Most of the syntax
Lua Version 5.4

C

The exact grammar is that of C11 with extensions implemented in popular compilers and additional extensions needed to allow processing of code with macros.

Note the following:

  • old K&R style of function declarations isn't parsed (there might be a workaround for it, but this syntax is deprecated either way)
  • preprocessor directives aren't tokenized according to language rules yet, but their contents is diffed
  • extensive use of macros in unusual places might not be parsed (this probably won't change)

Other than that code in C89, C99, C11 and GNU-versions of C language should be recognized.

C++

C++ support relies on external application called [srcml][srcml] and requires it to be installed in binary form (not necessary during build).

Reported standard version supported by srcml is C++14, so all previous ones should work too. Although their parser doesn't handle all language constructs equally well, it's seems to be good enough, especially for a ready to use parser that wasn't that hard to integrate.

Note the following:

  • the tuning of comparison is in progress and will be refined over time

Bash

The source of the grammar doesn't explicitly specify the version.

GNU Make

It's hard to measure level of support in case of GNU Make, because there seem to be no reference for the syntax itself apart from documentation, which is not concise.

Yet parser is capable of processing quite complicated examples of Makefiles (like the one used in this project or generated by automake) which contain many features most people don't know exist. It's definitely not 100%, but 90% or even more of all existing Makefiles out there should be parsed.

Note the following:

  • the comparison might not produce best results on Makefiles as it needs some tuning, this should happen over time (Makefiles aren't changed that often)

Lua

Note the following:

  • non-5.4 versions should work, but can produce worse results (however, syntax is normally backward compatible so this shouldn't happen often)

Other

More languages should be added in the future, maybe with external parsers that are capable of preserving all information about the source code.

CONFIGURATION

Configuration is done per directory tree ("root") which is the closes parent (or current directory) that contains .zs/ directory. The .zs/ directory contains files which define how contents of the root is to be processed. Settings from multiple nested roots are not combined.

.zs/exclude file

A .gitignore-like (or .git/info/exclude-like) file that lists paths relative to the root. The purpose is to exclude uninteresting files (automatically generated, third-party or otherwise). .zs/exclude is used by tools that search for files automatically and doesn't prevent the use of the same files when they are specified explicitly.

The following kinds of entries are recognized:

  • empty lines, which are ignored
  • lines that start with a # (comments), which are ignored
  • lines that end with / match only directories, the / is stripped and line processing continues
  • lines without / are treated as shell-like globs against filename which apply at any directory level and define paths whose processing should be skipped
  • lines that start with ! define exception from rules that precede them, you can't undo exclusion of files in excluded directories, for the purpose of this discussion the ! is stripped and line processing continues
  • lines that start with / always match paths instead of filename and provide a way to specify files to be ignored only in the root, otherwise they are processed as specified in the next item
  • other lines are treated as shell-like globs against paths relative to the root (leading / is allowed, but has no effect other than changing type of a match) which define paths whose processing should be skipped

No way to escape leading # and ! or a newline at the moment.

Globs support the following: [{char-class}], [!{char-class}], [^{char-class}], ? (doesn't match /), * (matches any (including zero) number of characters except for /) and \{char} (matches literal {char}).

Example:

# .zs/exclude

# automatically generated sources
src/c/c11-lexer.gen.cpp
src/c/c11-parser.gen.cpp
src/make/*.gen.*

# Qt-produced sources
ui_*.h
moc_*.cpp
moc_*.h

# file in root
/config.h

.zs/attributes file

Borrowing from the git project here again. This file consists of lines matching paths to attributes. Lines are trimmed before being processed.

Empty lines and comments work like in .zs/excludes file, all other lines follow this pattern:

exclude-expr [attr1=value1 [attr2=value2 [...]]]

Expressions that define exceptions (start with !) are recognized but ignored to keep syntax consistent between different files, which basically makes them another type of comments.

Each line of the file is visited in top down order and attributes from every matching entry are merged with the current state. Hierarchy of configuration values:

  1. Default values (lowest priority)
  2. Attributes
  3. Command-line parameters (highest priority)

Supported attributes:

  • lang
    Default: ""
    Those accepted by --lang command-line option: bash, c, cxx, make, lua
  • tab-size
    Default: 4
    Value should be an integer that's greater than zero

Unknown attributes are ignored.

Example:

# .zs/attributes

*.c tab-size=8
*.h tab-size=8 lang=c
tab-2.[ch] tab-size=2

# any.c has tab-size=8
# tab-2.c has tab-size=2
# tab-2.h has tab-size=2 lang=c
# any.h has tab-size=8 lang=c
# any.cpp has tab-size=4

INVOCATION

All tools have common and specific command-line arguments. Tool-specific arguments are described on the page of the tool, common ones are below.

-h, --help
print help message

--dry-run
parse and exit

--debug[=g]
enable debugging of grammar (g flag) and/or stree (s flag)

--dump[=t]
display tree (t flag) and/or stree (s flag)

--time-report
report time spent on different activities

--color
force colorization of output

--lang name
force specific language (bash, c, cxx, make, lua) and disable auto-detection

--no-pager
never spawn a pager for output

BEHAVIOUR

Pager

By default when output is a terminal (not the case when invoked by Git), less is spawn if output is about the size of the screen. Use --no-pager to change the behaviour.

Language detection

By default language is auto-detected based on file name. When name is not recognized the file is parsed as C.

TOOLS

zs-diff(1)

A terminal-based syntax-aware diff.

zs-find(1)

Grep-like tool that finds elements of source code structure.

zs-gdiff(1)

A Qt5 GUI version of syntax-aware diff.

zs-hi(1)

Simple syntax highlighter for xterm-256color palette.

zs-stats(1)

Counter of lines of code.

zs-tui(1)

TUI interface with underdefined scope of functionality.