Add support for bytes literals (#733)

Fixes #732

Author: chrisrink10

CHANGELOG.md

@@ -6,22 +6,23 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 ### Added
- * Added rudimentary support for `clojure.stacktrace` with `print-cause-trace` (part of #721).
+ * Added rudimentary support for `clojure.stacktrace` with `print-cause-trace` (part of #721)
+ * Added support for `bytes` literals using a `#b` prefix (#732)
 
 ### Fixed
- * Fix issue with `case` evaluating all of its clauses expressions (#699).
- * Fix issue with relative paths dropping their first character on MS-Windows (#703).
- * Fix incompatibility with `(str nil)` returning "nil" (#706).
- * Fix `sort-by` support for maps and boolean comparator fns (#709).
- * Fix `sort` support for maps and boolean comparator fns (#711).
- * Fix `(is (= exp act))` should only evaluate its args once on failure (#712).
- * Fix issue with `with` failing with a traceback error when an exception is thrown (#714).
- * Fix issue with `sort-*` family of funtions returning an error on an empty seq (#716).
- * Fix issue with `intern` failing when used (#725). 
- * Fix issue with `ns` not being available after `in-ns` on the REPL (#718).
- * Fixed issue with import modules aliasing using ns eval (#719).
- * Fix issue with `ns-resolve` throwing an error on macros (#720).
- * Fix issue with py module `readerwritelock` locks handling (#722).
+ * Fix issue with `case` evaluating all of its clauses expressions (#699)
+ * Fix issue with relative paths dropping their first character on MS-Windows (#703)
+ * Fix incompatibility with `(str nil)` returning "nil" (#706)
+ * Fix `sort-by` support for maps and boolean comparator fns (#709)
+ * Fix `sort` support for maps and boolean comparator fns (#711)
+ * Fix `(is (= exp act))` should only evaluate its args once on failure (#712)
+ * Fix issue with `with` failing with a traceback error when an exception is thrown (#714)
+ * Fix issue with `sort-*` family of funtions returning an error on an empty seq (#716)
+ * Fix issue with `intern` failing when used (#725)
+ * Fix issue with `ns` not being available after `in-ns` on the REPL (#718)
+ * Fixed issue with import modules aliasing using ns eval (#719)
+ * Fix issue with `ns-resolve` throwing an error on macros (#720)
+ * Fix issue with py module `readerwritelock` locks handling (#722)
 
 ## [v0.1.0a2]
 ### Added

docs/reader.rst

@@ -96,6 +96,38 @@ Strings are denoted as a series of characters enclosed by ``"`` quotation marks.
 If a string needs to contain a quotation mark literal, that quotation mark should be escaped as ``\"``.
 Strings may be multi-line by default and only a closing ``"`` will terminate reading a string.
 Strings correspond to the Python ``str`` type.
+String literals are always read with the UTF-8 encoding.
+
+String literals may contain the following escape sequences: ``\\``, ``\a``, ``\b``, ``\f``, ``\n``, ``\r``, ``\t``, ``\v``.
+Their meanings match the equivalent escape sequences supported in `Python string literals <https://docs.python.org/3/reference/lexical_analysis.html#string-and-bytes-literals>`_\.
+
+
+Byte Strings
+------------
+
+::
+
+    basilisp.user=> #b ""
+    #b ""
+    basilisp.user=> #b "this is a string"
+    #b "this is a string"
+    basilisp.user=> (type #b "")
+    <class 'bytes'>
+
+Byte strings are denoted as a series of ASCII characters enclosed by ``"`` quotation marks and preceded by a ``#b``.
+If a string needs to contain a quotation mark literal, that quotation mark should be escaped as ``\"``.
+Strings may be multi-line by default and only a closing ``"`` will terminate reading a string.
+Strings correspond to the Python ``bytes`` type.
+
+Byte string literals may contain the following escape sequences: ``\\``, ``\a``, ``\b``, ``\f``, ``\n``, ``\r``, ``\t``, ``\v``.
+Byte strings may also characters using a hex escape code as ``\xhh`` where ``hh`` is a hexadecimal value.
+Their meanings match the equivalent escape sequences supported in `Python byte string literals <https://docs.python.org/3/reference/lexical_analysis.html#string-and-bytes-literals>`_\.
+
+
+.. warning::
+
+   As in Python, byte string literals may not include any characters outside of the ASCII range.
+
 
 .. _character_literals:
 

src/basilisp/core.lpy

@@ -1321,11 +1321,6 @@
   [x]
   (and (integer? x) (neg? x)))
 
-(defn  ^:inline nil?
-  "Return ``true`` if ``x`` is ``nil``\\, otherwise ``false``\\."
-  [x]
-  (operator/is- x nil))
-
 (defn ^:inline some?
   "Return ``true`` if ``x`` is not ``nil``\\, otherwise ``false`` s."
   [x]
@@ -3603,7 +3598,7 @@
         ;;  pairs - pairs of bindings; either symbol/seq pairs or modifier pairs
         gen-iter (fn gen-iter [pairs]
                    (if (seq pairs)
-                     (let [for-iter (gensym "for") 
+                     (let [for-iter (gensym "for")
                            seq-arg  (gensym "seq")
                            pair     (first pairs)
                            binding  (first pair)

src/basilisp/lang/compiler/analyzer.py

@@ -3583,6 +3583,7 @@ def _const_node_type(_: Any) -> ConstType:
 
 for tp, const_type in {
     bool: ConstType.BOOL,
+    bytes: ConstType.BYTES,
     complex: ConstType.NUMBER,
     datetime: ConstType.INST,
     Decimal: ConstType.DECIMAL,
@@ -3612,6 +3613,7 @@ def _const_node_type(_: Any) -> ConstType:
 
 
 @_analyze_form.register(bool)
+@_analyze_form.register(bytes)
 @_analyze_form.register(complex)
 @_analyze_form.register(datetime)
 @_analyze_form.register(Decimal)
@@ -3649,6 +3651,7 @@ def _const_node(form: ReaderForm, ctx: AnalyzerContext) -> Const:
             form,
             (
                 bool,
+                bytes,
                 complex,
                 datetime,
                 Decimal,

src/basilisp/lang/compiler/generator.py

@@ -3275,6 +3275,7 @@ def _const_meta_kwargs_ast(
 
 
 @_const_val_to_py_ast.register(bool)
+@_const_val_to_py_ast.register(bytes)
 @_const_val_to_py_ast.register(type(None))
 @_const_val_to_py_ast.register(complex)
 @_const_val_to_py_ast.register(float)

src/basilisp/lang/compiler/nodes.py

@@ -271,6 +271,7 @@ class ConstType(Enum):
     SET = kw.keyword("set")
     VECTOR = kw.keyword("vector")
     BOOL = kw.keyword("bool")
+    BYTES = kw.keyword("bytes")
     KEYWORD = kw.keyword("keyword")
     SYMBOL = kw.keyword("symbol")
     STRING = kw.keyword("string")

src/basilisp/lang/obj.py

@@ -202,6 +202,12 @@ def _lrepr_bool(o: bool, **_) -> str:
     return repr(o).lower()
 
 
+@lrepr.register(bytes)
+def _lrepr_bytes(o: bytes, **_) -> str:
+    v = repr(o)
+    return f'#b "{v[2:-1]}"'
+
+
 @lrepr.register(type(None))
 def _lrepr_nil(_: None, **__) -> str:
     return "nil"

src/basilisp/lang/reader.py

@@ -843,7 +843,7 @@ def _read_num(  # noqa: C901  # pylint: disable=too-many-statements
 
 
 def _read_str(ctx: ReaderContext, allow_arbitrary_escapes: bool = False) -> str:
-    """Return a string from the input stream.
+    """Return a UTF-8 encoded string from the input stream.
 
     If allow_arbitrary_escapes is True, do not throw a SyntaxError if an
     unknown escape sequence is encountered."""
@@ -869,6 +869,75 @@ def _read_str(ctx: ReaderContext, allow_arbitrary_escapes: bool = False) -> str:
         s.append(token)
 
 
+_BYTES_ESCAPE_CHARS = {
+    '"': b'"',
+    "\\": b"\\",
+    "a": b"\a",
+    "b": b"\b",
+    "f": b"\f",
+    "n": b"\n",
+    "r": b"\r",
+    "t": b"\t",
+    "v": b"\v",
+}
+
+
+def _read_hex_byte(ctx: ReaderContext) -> bytes:
+    """Read a byte with a 2 digit hex code such as `\\xff`."""
+    reader = ctx.reader
+    c1 = reader.next_token()
+    c2 = reader.next_token()
+    try:
+        return bytes([int("".join(["0x", c1, c2]), base=16)])
+    except ValueError as e:
+        raise ctx.syntax_error(
+            f"Invalid byte representation for base 16: 0x{c1}{c2}"
+        ) from e
+
+
+def _read_byte_str(ctx: ReaderContext) -> bytes:
+    """Return a byte string from the input stream.
+
+    Byte strings have the same restrictions and semantics as byte literals in Python.
+    Individual characters must be within the ASCII range or must be valid escape sequences.
+    """
+    reader = ctx.reader
+
+    token = reader.peek()
+    while whitespace_chars.match(token):
+        token = reader.next_token()
+
+    if token != '"':
+        raise ctx.syntax_error(f"Expected '\"'; got '{token}' instead")
+
+    b: List[bytes] = []
+    while True:
+        token = reader.next_token()
+        if token == "":
+            raise ctx.eof_error("Unexpected EOF in byte string")
+        if ord(token) < 1 or ord(token) > 127:
+            raise ctx.eof_error("Byte strings must contain only ASCII characters")
+        if token == "\\":
+            token = reader.next_token()
+            escape_char = _BYTES_ESCAPE_CHARS.get(token, None)
+            if escape_char:
+                b.append(escape_char)
+                continue
+            elif token == "x":
+                b.append(_read_hex_byte(ctx))
+                continue
+            else:
+                # In Python, invalid escape sequences entered into byte strings are
+                # retained with backslash for debugging purposes, so we do the same.
+                b.append(b"\\")
+                b.append(token.encode("utf-8"))
+                continue
+        if token == '"':
+            reader.next_token()
+            return b"".join(b)
+        b.append(token.encode("utf-8"))
+
+
 @_with_loc
 def _read_sym(ctx: ReaderContext) -> MaybeSymbol:
     """Return a symbol from the input stream.
@@ -1380,7 +1449,7 @@ def _load_record_or_type(
         raise ctx.syntax_error("Records may only be constructed from Vectors and Maps")
 
 
-def _read_reader_macro(ctx: ReaderContext) -> LispReaderForm:
+def _read_reader_macro(ctx: ReaderContext) -> LispReaderForm:  # noqa: MC0001
     """Return a data structure evaluated as a reader
     macro from the input stream."""
     start = ctx.reader.advance()
@@ -1408,6 +1477,9 @@ def _read_reader_macro(ctx: ReaderContext) -> LispReaderForm:
         return _read_reader_conditional(ctx)
     elif token == "#":
         return _read_numeric_constant(ctx)
+    elif token == "b":
+        ctx.reader.advance()
+        return _read_byte_str(ctx)
     elif ns_name_chars.match(token):
         s = _read_sym(ctx)
         assert isinstance(s, sym.Symbol)

src/basilisp/lang/typing.py

@@ -21,6 +21,7 @@
 LispNumber = Union[int, float, Fraction]
 LispForm = Union[
     bool,
+    bytes,
     complex,
     datetime,
     Decimal,

tests/basilisp/lrepr_test.py

@@ -173,6 +173,11 @@ def test_print_readably(lcompile: CompileFn):
             '#uuid "81f35603-0408-4b3d-bbc0-462e3702747f"',
             '(pr-str #uuid "81f35603-0408-4b3d-bbc0-462e3702747f")',
         ),
+        ('#b ""', '(pr-str #b "")'),
+        (
+            r'#b "\x7fELF\x01\x01\x01\x00"',
+            r'(pr-str #b "\x7f\x45\x4c\x46\x01\x01\x01\x00")',
+        ),
         ('#"\\s"', '(pr-str #"\\s")'),
         (
             '#inst "2018-11-28T12:43:25.477000+00:00"',
@@ -205,6 +210,11 @@ def test_lrepr(lcompile: CompileFn, repr: str, code: str):
         (-float("inf"), "(read-string (pr-str ##-Inf))"),
         ("hi", '(read-string (pr-str "hi"))'),
         ("Hello\nworld!", '(read-string (pr-str "Hello\nworld!"))'),
+        (b"", '(read-string (pr-str #b ""))'),
+        (
+            b"\x7fELF\x01\x01\x01\x00",
+            r'(read-string (pr-str #b "\x7f\x45\x4c\x46\x01\x01\x01\x00"))',
+        ),
         (
             uuid.UUID("81f35603-0408-4b3d-bbc0-462e3702747f"),
             '(read-string (pr-str #uuid "81f35603-0408-4b3d-bbc0-462e3702747f"))',
@@ -253,6 +263,11 @@ def test_lrepr_round_trip_special_cases(lcompile: CompileFn):
         ("##-Inf", "(print-str ##-Inf)"),
         ("hi", '(print-str "hi")'),
         ("Hello\nworld!", '(print-str "Hello\nworld!")'),
+        ('#b ""', '(print-str #b "")'),
+        (
+            r'#b "\x7fELF\x01\x01\x01\x00"',
+            r'(print-str #b "\x7f\x45\x4c\x46\x01\x01\x01\x00")',
+        ),
         # In Clojure, (print-str #uuid "...") produces '#uuid "..."' but in Basilisp
         # print-str is tied directly to str (which in Clojure simply returns the string
         # part of the UUID).