circuitpython/tools/tinytest-codegen.py
Jim Mussared 692d36d779 py: Implement partial PEP-498 (f-string) support.
This implements (most of) the PEP-498 spec for f-strings and is based on
https://github.com/micropython/micropython/pull/4998 by @klardotsh.

It is implemented in the lexer as a syntax translation to `str.format`:
  f"{a}" --> "{}".format(a)

It also supports:
  f"{a=}" --> "a={}".format(a)

This is done by extracting the arguments into a temporary vstr buffer,
then after the string has been tokenized, the lexer input queue is saved
and the contents of the temporary vstr buffer are injected into the lexer
instead.

There are four main limitations:
- raw f-strings (`fr` or `rf` prefixes) are not supported and will raise
  `SyntaxError: raw f-strings are not supported`.

- literal concatenation of f-strings with adjacent strings will fail
    "{}" f"{a}" --> "{}{}".format(a)    (str.format will incorrectly use
                                         the braces from the non-f-string)
    f"{a}" f"{a}" --> "{}".format(a) "{}".format(a) (cannot concatenate)

- PEP-498 requires the full parser to understand the interpolated
  argument, however because this entirely runs in the lexer it cannot
  resolve nested braces in expressions like
    f"{'}'}"

- The !r, !s, and !a conversions are not supported.

Includes tests and cpydiffs.

Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
2021-08-14 16:58:40 +10:00

136 lines
4.1 KiB
Python
Executable File

#!/usr/bin/env python3
import os, sys
from glob import glob
from re import sub
import argparse
def escape(s):
s = s.decode()
lookup = {
"\0": "\\0",
"\t": "\\t",
"\n": '\\n"\n"',
"\r": "\\r",
"\\": "\\\\",
'"': '\\"',
}
return '""\n"{}"'.format("".join([lookup[x] if x in lookup else x for x in s]))
def chew_filename(t):
return {"func": "test_{}_fn".format(sub(r"/|\.|-", "_", t)), "desc": t}
def script_to_map(test_file):
r = {"name": chew_filename(test_file)["func"]}
with open(test_file, "rb") as f:
r["script"] = escape(f.read())
with open(test_file + ".exp", "rb") as f:
r["output"] = escape(f.read())
return r
test_function = (
"void {name}(void* data) {{\n"
" static const char pystr[] = {script};\n"
" static const char exp[] = {output};\n"
' printf("\\n");\n'
" upytest_set_expected_output(exp, sizeof(exp) - 1);\n"
" upytest_execute_test(pystr);\n"
' printf("result: ");\n'
"}}"
)
testcase_struct = "struct testcase_t {name}_tests[] = {{\n{body}\n END_OF_TESTCASES\n}};"
testcase_member = ' {{ "{desc}", {func}, TT_ENABLED_, 0, 0 }},'
testgroup_struct = "struct testgroup_t groups[] = {{\n{body}\n END_OF_GROUPS\n}};"
testgroup_member = ' {{ "{name}", {name}_tests }},'
## XXX: may be we could have `--without <groups>` argument...
# currently these tests are selected because they pass on qemu-arm
test_dirs = (
"basics",
"micropython",
"misc",
"extmod",
"float",
"inlineasm",
"qemu-arm",
) # 'import', 'io',)
exclude_tests = (
# pattern matching in .exp
"basics/bytes_compare3.py",
"extmod/ticks_diff.py",
"extmod/time_ms_us.py",
"extmod/uheapq_timeq.py",
# unicode char issue
"extmod/ujson_loads.py",
# doesn't output to python stdout
"extmod/ure_debug.py",
"extmod/vfs_basic.py",
"extmod/vfs_fat_ramdisk.py",
"extmod/vfs_fat_fileio.py",
"extmod/vfs_fat_fsusermount.py",
"extmod/vfs_fat_oldproto.py",
# rounding issues
"float/float_divmod.py",
# requires double precision floating point to work
"float/float2int_doubleprec_intbig.py",
"float/float_parse_doubleprec.py",
# inline asm FP tests (require Cortex-M4)
"inlineasm/asmfpaddsub.py",
"inlineasm/asmfpcmp.py",
"inlineasm/asmfpldrstr.py",
"inlineasm/asmfpmuldiv.py",
"inlineasm/asmfpsqrt.py",
# different filename in output
"micropython/emg_exc.py",
"micropython/heapalloc_traceback.py",
# don't have emergency exception buffer
"micropython/heapalloc_exc_compressed_emg_exc.py",
# pattern matching in .exp
"micropython/meminfo.py",
# needs sys stdfiles
"misc/print_exception.py",
# settrace .exp files are too large
"misc/sys_settrace_loop.py",
"misc/sys_settrace_generator.py",
"misc/sys_settrace_features.py",
# don't have f-string
"basics/string_fstring.py",
)
output = []
tests = []
argparser = argparse.ArgumentParser(
description="Convert native MicroPython tests to tinytest/upytesthelper C code"
)
argparser.add_argument("--stdin", action="store_true", help="read list of tests from stdin")
argparser.add_argument("--exclude", action="append", help="exclude test by name")
args = argparser.parse_args()
if not args.stdin:
if args.exclude:
exclude_tests += tuple(args.exclude)
for group in test_dirs:
tests += [test for test in glob("{}/*.py".format(group)) if test not in exclude_tests]
else:
for l in sys.stdin:
tests.append(l.rstrip())
output.extend([test_function.format(**script_to_map(test)) for test in tests])
testcase_members = [testcase_member.format(**chew_filename(test)) for test in tests]
output.append(testcase_struct.format(name="", body="\n".join(testcase_members)))
testgroup_members = [testgroup_member.format(name=group) for group in [""]]
output.append(testgroup_struct.format(body="\n".join(testgroup_members)))
## XXX: may be we could have `--output <filename>` argument...
# Don't depend on what system locale is set, use utf8 encoding.
sys.stdout.buffer.write("\n\n".join(output).encode("utf8"))