Tree @master (Download .tar.gz)

HISTORY.md @master — view markup · raw · history · blame

No changes to language other than to add tests to the spec that emphasize that quoted atoms may contain UTF-8-encoded Unicode characters.
Converted Python source files from Python 2.7 to Python 3.

The main thing here beyond the usual 2-to-3 stuff was dealing with Tamsin's use of strings, which are both used heavily in the elaborate Scanner/ScannerEngine setup and tied strongly to Python 2's string model.

The solution I went with was to treat Python 3 strs as "8-bit-clean" strings, encoded as latin-1 for the sake of being able to treat them like lists of bytes, but still able to contain Unicode characters, as UTF-8 encoded byte sequences.

An alternate solution would be to retain them as Python's binary type. This might've been a better solution. * Fixed bugs in tamsin parse and tamsin desugar subcommands. * Updated lib/mini-tamsin.tamsin so that it at least parses with the parser-in-Tamsin. Added comments identifying issues.

No changes to language. (Really, if we were trying to be closer to SemVer, then 0.5-2017.0502 should've been 0.6, and 0.5-2026.0804 would've been 0.6-2026.0804, and this would've been another 0.6-2026.xxxx, but alas it is too late for that now.)
Fixed obvious bug in tamsin scan subcommand.
Updated some broken test expectations for micro-tamsin.
Renamed .markdown files to .md.
Arranged licensing info according to the REUSE convention, version 3.2.

No changes to language.
Pin python2.7 as the explicit binary for Python, in scripts. (Just python is often ambiguous or missing on systems.)
Tiny repository cleanups.

This is an interim release, created because the tests pass here, even though not everything aimed for for the next release has been achieved.

mini-tamsin.tamsin is an interpreter for "Mini-Tamsin", written in Tamsin.
Better error reporting.
Improvements or bugfixes in the C-language implementation of $:unquote.
Tamsin programs can handle streams on input and produce streams on output.
Begun work on a better C-emitting backend.
Better scanning; buffers are more sophisticated and track some state themselves.

EOF is no longer a special kind of term; it is no longer exposed, as a value, to Tamsin programs. ($:eof returns '' on success.)
Prolog/Erlang-style list sugar for terms, in patterns as well.
When a new scanner is switched to using using, that scanner defaults to the $:utf8 scanner for its scanning. This prevents the common shooting-self-in-foot error of selecting a production that is not itself using another scanner (which would result in an infinite loop of the production scanner trying to use itself as its subsidiary scanner.)

struct term *s are (almost) always const in compiled Tamsin programs (for better sharing; we don't need to make copies of them)
related: variable-matching is more efficient (directly updates an array of terms, instead of searching for the variable by name)
related: creating new atoms uses hash-consing, so that no new struct term for the atom is allocated if one already exists (the existing one is shared.) This reduces memory usage significantly.

Tamsin-to-C compiler written in Tamsin (mains/compiler.tamsin) passes all tests, and can compile itself.
Refactored $ functions into tamsin.sysmod module in Python version.

$:equal now does deep equality of arbitrary ground terms.
$:repr added.
$:reverse added.
Some standard modules ship in the distribution: list, tamsin_scanner, and tamsin_parser.

Support for user-defined modules.
tamsin can take more than one source file on command line; this is how external modules are supported (by this implementation.)
Cleaned-up testing framework; Tamsin versions of scanner, grammar, parser, desugarer, analyzer, and compiler found in mains subdir.
Most tamsin verbs, and their versions in Tamsin, corresponding to intermediate phases, output reprified terms.
tamsin significantly re-factored so that the interpreter and compiler are more similar, and generating code for production branches is easier.
Added Tamsin-to-C compiler written in Tamsin, which can pass the first 43 or so tests from the spec ("Mini-Tamsin").

Module-member syntax changed from . to :.
: can be used without any module on the LHS to refer to a production in the current module.
Added "fold" forms, binary / and ternary //.

$:char scanner dropped. Instead, there are $:byte (which always returns 8-bit-clean bytes) and $:utf8 (which always returns UTF-8 sequences.)
Added $:equal(L,R).
$:unquote(X,L,R) takes three arguments now.

Beginnings of user-defined module support (very rudimentary, not to be used.)
Code in libtamsin is much more robust. AST-builder written in Tamsin now compiles and runs correctly.
Added a desugaring phase to tamsin, and a desugarer written in Tamsin.
Added Micro-Tamsin interpreter, written in Tamsin. Can pass the first 30 tests from the spec.

Initial release.