bp/bp.1.md

% BP(1)
% Bruce Hill (*bruce@bruce-hill.com*)
% May 17 2021

# NAME

bp - Bruce\'s Parsing Expression Grammar tool

# SYNOPSIS

`bp` \[*options...*\] *pattern* \[\[`--`\] *files...*\]

# DESCRIPTION

`bp` is a tool that matches parsing expression grammars using a custom
syntax.

# OPTIONS

`-p`, `--pattern` *pat*
: Give a pattern in BP syntax instead of string syntax (equivalent to `bp
'\(pat)'`

`-w`, `--word` *word*
: Surround a string pattern with word boundaries (equivalent to `bp
'\|word\|'`)

`-e`, `--explain`
: Print a visual explanation of the matches.

`-j`, `--json`
: Print a JSON list of the matches. (Pairs with `--verbose` for more detail)

`-l`, `--list-files`
: Print only the names of files containing matches instead of the matches
themselves.

`-i`, `--ignore-case`
: Perform pattern matching case-insensitively.

`-I`, `--inplace`
: Perform filtering or replacement in-place (i.e. overwrite files with new
content).

`-r`, `--replace` *replacement*
: Replace all occurrences of the main pattern with the given string.

`-s`, `--skip` *pattern*
: While looking for matches, skip over *pattern* occurrences. This can be
useful for behavior like `bp -s string` (avoiding matches inside string
literals).

`-g`, `--grammar` *grammar-file*
: Load the grammar from the given file. See the `GRAMMAR FILES` section
for more info.

`-G`, `--git`
: Use `git` to get a list of files. Remaining file arguments (if any) are
passed to `git --ls-files` instead of treated as literal files.

`-B`, `--context-before` *N*
: The number of lines of context to print before each match (default: 0). See
`--context` below for details on `none` or `all`.

`-A`, `--context-after` *N*
: The number of lines of context to print after each match (default: 0). See
`--context` below for details on `none` or `all`.


`-C`, `--context` *N*
: The number of lines to print before and after each match (default: 0). If *N*
is `none`, print only the exact text of the matches. If *N* is **"all"**, print
all text before and after each match.

`-f`, `--format` *fancy*\|*plain*\|*bare*\|*file:line*\|*auto*
: Set the output format. *fancy* includes colors and line numbers, *plain*
prints line numbers with no coloring, *bare* prints only the match text,
*file:line* prints the filename and line number for each match (grep-style),
and *auto* (the default) uses *fancy* formatting when the output is a TTY and
*bare* formatting otherwise.

`-h`, `--help`
: Print the usage and exit.

*pattern*
: The main pattern for bp to match. By default, this pattern is a string
pattern (see the **STRING PATTERNS** section below).

*files...*
: The input files to search. If no input files are provided and data was piped
in, that data will be used instead. If neither are provided, `bp` will search
through all files in the current directory and its subdirectories
(recursively).


# STRING PATTERNS

One of the most common use cases for pattern matching tools is matching plain,
literal strings, or strings that are primarily plain strings, with one or two
patterns. `bp` is designed around this fact. The default mode for bp patterns
is "string pattern mode". In string pattern mode, all characters are
interpreted literally except for the backslash (`\`), which may be followed by
an escape or a bp pattern (see the **PATTERNS** section below). Optionally, the
bp pattern may be terminated by a semicolon (`;`).


# PATTERNS

`bp` patterns are based off of a combination of Parsing Expression Grammars and
regular expression syntax. The syntax is designed to map closely to verbal
descriptions of the patterns, and prefix operators are preferred over suffix
operators (as is common in regex syntax). Patterns are whitespace-agnostic, so
they work the same regardless of whether whitespace is present or not, except
for string literals (`'...'` and `"..."`), character literals (`` ` ``), and
escape sequences (`\`). Whitespace between patterns or parts of a pattern
should be used for clarity, but it will not affect the meaning of the pattern.

*pat1 pat2*
: A sequence: *pat1* followed by *pat2*

*pat1* `/` *pat2*
: A choice: *pat1*, or if it doesn\'t match, then *pat2*

`.`
: Any character (excluding newline)

`^`
: Start of a line

`^^`
: Start of the text

`$`
: End of a line (does not include newline character)

`$$`
: End of the text

`_`
: Zero or more whitespace characters, including spaces and tabs, but not
newlines.

`__`
: Zero or more whitespace characters, including spaces, tabs, newlines, and
comments. Comments are undefined by default, but may be defined by a separate
grammar file. See the **GRAMMAR FILES** section for more info.

`"foo"`, `'foo'`
: The literal string **"foo"**. Single and double quotes are treated the same.
Escape sequences are not allowed.

`` ` ``*c*
: The literal character *c* (e.g. `` `@ `` matches the "@" character)

`` ` ``*c1*`-`*c2*
: The character range *c1* to *c2* (e.g. `` `a-z ``). Multiple ranges
can be combined with a comma (e.g. `` `a-z,A-Z ``).

`` ` ``*c1*`,`*c2*
: Any one of the given character or character ranges *c1* or *c2* (e.g. `` `a,e,i,o,u,0-9 ``)

`\`*esc*
: An escape sequence (e.g. `\n`, `\x1F`, `\033`, etc.)

`\`*esc1*`-`*esc2*
: An escape sequence range from *esc1* to *esc2* (e.g. `\x00-x1F`)

`\`*esc1*`,`*esc2*
: Any one of the given escape sequences or ranges *esc1* or *esc2* (e.g. `\r,n,x01-x04`)

`\N`
: A special escape that matches a "nodent": one or more newlines followed by
the same indentation that occurs on the current line.

`\C`
: A special escape that always matches the empty string and replaces it with
the indentation of the line on which it matched. For example, this pattern
would match Bash-style heredocs that start with "<<-FOO" and end with a line
containing only the starting indentation and the string "FOO":
`"<<-" @end=(\C id) ..%\n (^end$)`

`\i`
: An identifier character (e.g. alphanumeric characters or underscores).

`\I`
: An identifier character, not including numbers (e.g. alphabetic characters or underscores).

`|`
: A word boundary (i.e. the edge of a word).

`\b`
: Alias for `|` (word boundary)

`!` *pat*
: Not *pat*

`[` *pat* `]`
: Maybe *pat*

*N* *pat*
: Exactly *N* repetitions of *pat* (e.g. `5 "x"` matches **"xxxxx"**)

*N* `-` *M* *pat*
: Between *N* and *M* repetitions of *pat* (e.g. `2-3 "x"` matches **"xx"** or
**"xxx"**)

*N*`+` *pat*
: At least *N* or more repetitions of *pat* (e.g. `2+ "x"` matches
**"xx"**, **"xxx"**, **"xxxx"**, etc.)

`*` *pat*
: Any *pat*s (zero or more, e.g. `* "x"` matches **""**, **"x"**, **"xx"**,
etc.)

`+` *pat*
: Some *pat*s (e.g. `+ "x"` matches **"x"**, **"xx"**, **"xxx"**, etc.)

*repeating-pat* `%` *sep*
: *repeating-pat* (see the examples above) separated by *sep* (e.g. `*word %
","` matches zero or more comma-separated words)

`..` *pat*
: Any text (except newlines) up to and including *pat*

`.. %` *skip* *pat*
: Any text (except newlines) up to and including *pat*, skipping over instances
of *skip* (e.g. `'"' ..%('\' .) '"'` opening quote, up to closing quote,
skipping over backslash followed by a single character)

`.. =` *only* *pat*
: Any number of repetitions of the pattern *only* up to and including *pat*
(e.g. `"f" ..=abc "k"` matches the letter "f" followed by some alphabetic
characters and then a "k", which would match "fork", but not "free kit") This
is essentially a "non-greedy" version of `*`, and `.. pat` can be thought of as
the special case of `..=. pat`

`<` *pat*
: Matches at the current position if *pat* matches immediately before the
current position (lookbehind). Conceptually, you can think of this as creating
a file containing only the *N* characters immediately before the current
position and attempting to match *pat* on that file, for all values of *N* from
the minimum number of characters *pat* can match up to maximum number of
characters *pat* can match (or the length of the current line upto the current
position, whichever is smaller). **Note:** For fixed-length lookbehinds, this
is quite efficient (e.g. `<(100 "x")`), however this could cause performance
problems with variable-length lookbehinds (e.g. `<("x" 0-100"y")`). Also, it is
worth noting that `^`, `^^`, `$`, and `$$` all match against the edges of the
slice, which may give false positives if you were expecting them to match only
against the edges file or line.

`>` *pat*
: Matches *pat*, but does not consume any input (lookahead).

`@` *pat*
: Capture *pat*

`foo`
: The named pattern whose name is **"foo"**. Pattern names come from
definitions in grammar files or from named captures. Pattern names may contain
dashes (`-`), but not underscores (`_`), since the underscore is used to match
whitespace. See the **GRAMMAR FILES** section for more info.

`@` *name* `=` *pat*
: Let *name* equal *pat* (named capture). Named captures can be used as
backreferences like so: `` @foo=word `( foo `) `` (matches **"asdf(asdf)"** or
**"baz(baz)"**, but not **"foo(baz)"**)

*pat* `=>` `"`*replacement*`"`
: Replace *pat* with *replacement*. Note: *replacement* should be a string
(single or double quoted), and it may contain escape sequences (e.g. `\n`) or
references to captured values: `@0` (the whole of *pat*), `@1` (the first
capture in *pat*), `@`*foo* (the capture named *foo* in *pat*), etc. For
example, `@word _ @rest=(*word % _) => "@rest:\n\t@1"` matches a word followed
by whitespace, followed by a series of words and replaces it with the series
of words, a colon, a newline, a tab, and then the first word.

*pat1* `~` *pat2*
: Matches when *pat1* matches and *pat2* can be found within the text of that
match. (e.g. `comment ~ {TODO}` matches comments that contain the word
**"TODO"**)

*pat1* `!~` *pat2*
: Matches when *pat1* matches, but *pat2* can not be found within the text of
that match. (e.g. `comment ~ {IGNORE}` matches only comments that do not
contain the word **"IGNORE"**)

*name*`:` *pat*
: Define *name* to mean *pat* (pattern definition)

*name*`::` *pat*
: Define *name* to be a special tagged pattern *pat*. This is the same as a
regular definition, except that a piece of metadata is attached to it
associating it with the specified name.

`#` *comment*
: A line comment


# GRAMMAR FILES

**bp** allows loading extra grammar files, which define patterns which may be
used for matching. The **builtins** grammar file is loaded by default, and it
defines a few useful general-purpose patterns. For example, it defines the
**parens** rule, which matches pairs of matching parentheses, accounting for
nested inner parentheses:

```
bp -p '"my_func" parens'
```

**bp** also comes with a few grammar files for common programming languages,
which may be loaded on demand. These grammar files are not comprehensive syntax
definitions, but only some common patterns. For example, the c++ grammar file
contains definitions for `//`-style line comments as well as `/*...*/`-style
block comments. Thus, you can find all comments with the word "TODO" with the
following command:

```
bp -g c++ -p 'comment ~ {TODO}' *.cpp
```


# EXAMPLES

Find files containing the string "foo" (a string pattern):
```
ls | bp foo
```

Find files ending with ".c" and print the name with the ".c" replaced with ".h":
```
ls | bp '.c\$' -r '.h'
```

Find the word "foobar", followed by a pair of matching parentheses in the file
*my_file.py*:
```
bp -p '{foobar} parens' my_file.py
```

Using the *html* grammar, find all *element*s matching the tag *a* in the file
*foo.html*:
```
bp -g html -p 'element ~ (^^"<a ")' foo.html
```