New features:

• implement structured pipelines and document them
• add shell builtins answers dir first from parse proc select skip sort-by to where that read and/or write structured data
• add buffering to (console-input-port) (console-output-port) and (console-error-port)
• add functions (dir) (dirs) (file) (files)
• and type process-entry and function (proc) to list active processes
• add functions (sh-start/reader1) (tty-colors) and macro (with-sh-closable)
• add types reader and writer
• add many functions for reader types: (all) (all/vector) (close) (constant-reader) (copy-all) (copy-all/close) (datum-reader) (empty-reader) (eof?) (first) (get) (in-reader) (iterator-reader) (list-reader) (put) (reader->list) (reader->vector) (readers) (skip) (skip-first) (vector-reader)
• add some functions for writer types: (discard-writer) (full-writer) (list-writer) (vector-writer)
• add several functions (from-...) and (to-...)
• add macros for reader and writer types: (for-reader) (filter==>) (select) (sort-by) (where) (where@)
• add types field-reader filter-reader name0-reader range-reader and sort-reader
• add types fifo-reader fifo-writer and new macro (thread==>) that uses them
• add types json-reader and json-writer
• add types queue-reader and queue-writer
• add type table-writer
• add types wire-reader and wire-writer
• add library (scheme2k reflect) with functions for simplifing reflection
• add function (wire-get-from-bytespan)

Improvements:

• pretty-print types time and date and add convenience functions to create and convert them
• delete macro (for) and rename macro (for*) -> (for)
• rename functions (...producer...) -> (...queue-reader...)
• rename functions (...consumer...) -> (...queue-writer...)
• rename function (...channel...) -> (...wire-reader...) or (...wire-writer...)
• rename library (schemesh ipc fifo) -> (scheme2k ipc queue)
• rename library (scheme2k wire) -> (scheme2k io wire) and extend it serialize/deserialize date objects and ordered-hash objects

New features:

• change license GPLv2+ -> LGPLv2+ in all sources except parser/* shell/* repl/* main.c test/*
• rename all LGPLv2+ C functions and LGPLv2+ Scheme libraries schemesh* -> scheme2k*
• and add optional Makefile targets scheme2k_so and scheme2k_c_so to compile LGPLv2+ sources to new libraries libscheme2k_0.9.3.so and libscheme2k_c_0.9.3.so that can be loaded from vanilla Chez Scheme.
• add optional Makefile target scheme2k_http_c_so, compiles LGPLv2+ library libscheme2k_http_c_0.9.3.so containing bindings for using C libcurl library from Chez Scheme (requires schemesh or libscheme2k)
• add shell builtin ulimit
• add macro (for-glob), iterates on filesystem paths generated by shell glob syntax
• add function (fd-nonblock?)
• add key bindings for more xterm keys

Improvements:

• improve documentation
• rename function (channel-fd) -> (channel) and extend it to also accept binary ports in alternative to file descriptors
• rename function (fd-setnonblock) -> (fd-nonblock?-set!)
• rename function (in-shell-glob) -> (in-glob)
• rename function (open-pipe-fds) -> (pipe-fds)
• rename function (open-socketpair-fds) -> (socketpair-fds)
• rename function (wildcard) -> (wildcard1+) and add function (wildcard)

Fixes:

• fix issue #33: compile error on macOS due to file descriptor number higher than undocumented maximum. Note: ulimit -n can be raised above the actually supported maximum on macOS, so its current value may be misleading.

New features:

• subsume functions (sh-binary-port) and (sh-textual-port) into the new function (sh-port) for accessing redirections of Scheme jobs
• add more SRFI 18 compatible functions
• add support for keywords while and until inside macros (for) (for*) (for-...) (forever) (repeat) (until) (while)
• add key bindings for more xterm keys
• export more linectx functions
• add more documentation, including: how to define new key bindings, how to add job redirections from Scheme syntax

Improvements:

• implement autocomplete for shell paths starting with ~ - supersedes pull request #29
• improve (sh-eval) to better handle forms containing (begin ...) followed by alternating functions and macros
• modify functions (file->port) and (sh-port) to return textual ports by default
• allow calling (sh-redirect) on running jobs, and document its behavior in such case

Fixes:

• fix issue #30: bug in (string-join)
• fix wildcards *FOO to also match directories whose name ends with FOO, instead of only matching non-directories
• fix an off-by-one error in wildcard expansion

New features:

• add functions (sh-start/fds) (sh-start/ports)
• add macro (for-alist) and function (in-alist)
• add shell builtin threads to list existing threads and their status
• add thread management functions (get-thread) (thread) (thread-alive?) (thread-count) (thread-find) (thread-id) (thread-initial-bindings) (thread-kill) (threads) (threads-status)

Fixes:

• fix saving history to ~/.cache/schemesh/history.txt
• fix issue #26: flush all open C FILE* streams before fork()ing a subshell

New features:

• add build instructions for Windows using WSL
• add build instructions for NixOS
• add functions (sh-binary-port) (sh-textual-port) for accessing redirections of Scheme jobs
• add functions (fork-process) (repl-answers) (repl-answers-display) (repl-history) (repl-history-display)
• add functions creating customized, colored prompt and document them
• add Racket-compatible functions (port->list) (port->string) (port->bytes) (port->lines) (port->bytes-lines)
(read-line) (read-bytes-line) (lines->port) (byte-lines->port)
• add macro (==>)

Improvements:

• improve shell builtins fg bg and wait: job-id is now optional
• improve autocompletion inside single or double quotes
• display colored jobs status
• add per-job ports, needed for buffering input/output of redirected Scheme jobs
• refactor type vscreen to contain colored chars: will allow implementing syntax highlighting
• install interruptible, unbuffered UTF-8b ports as (console-input-port) (console-output-port) and (console-error-port)
• continue implementing library (schemesh ipc fifo)

Fixes:

• fix job control bugs on macOS, closes issue #17
• delete shell builtin unsafe, closes issue #23
• fix bugs in replacement functions (get-bytevector-some) (putenv) (bytevector-sint-ref)
• fix signal handlers to correctly react to signals received by secondary threads

New features:

• add functions (channel?) (channel-fd) (channel-pipe-pair) (channel-socket-pair) (channel-close) (channel-get) (channel-eof?) (channel-put) (in-channel) for exchanging data between different processes through a socket, pipe or analogous file descriptor. They internally serialize and deserialize data using library (schemesh wire)
• add function (wire-register-rtd-fields) to quickly create serializers and deserializers for user-defined record types

Improvements:

• improve (sh-run/string) and (sh-run/bvector) behavior if job being executed receives a signal
• define optimized versions of standard - but currently slow - functions
  (bytevector-sint-ref) (bytevector-sint-set!) (bytevector-uint-ref)(bytevector-uint-set!)
• add support for serializing and deserializing Chez Scheme time object to library (schemesh wire)

Fixes:

• fix a visual glitch when pasting multiple lines at REPL, and last line does not end with a #\newline
• remove Apache-licensed file containers/hashtable-types.ss that was copy-pasted from Chez Scheme sources: replace it with accessor functions obtained from record-type-descriptor objects.

New features:

• add shell builtin wait
• add macros (shell-glob) and (shell-string) for expanding shell syntax to, respectively, a list of strings or a single string.
• add macro (in-shell-glob) as a shortcut for (in-list (shell-glob))
• rename function (sh-version) -> (sh-version-number) and add function (sh-version) that returns a string
• add functions (bytespan->bytevector0) (countdown) (chargbuffer->charspans*) (chargbuffer->string) (charlines->charspan) (charlines->string) (fd-seek) (open-socketpair-fds) and macro (with-raw-tty)
• add standalone executable countdown as an alternative to sleep
• add serialization/deserialization library (schemesh wire) that exports the functions (datum->wire) (wire->datum) (wire-get) (wire-length) (wire-put) and can also be extended to support user-defined record types

Improvements:

• extend functions (sh-find-job) and (sh-job) to accept #f as a shortcut for (or (sh-current-job) (sh-globals))
• improve functions (sh-fg) (sh-bg) (sh-wait) to re-raise exceptions thrown by Scheme jobs
• add lots of functions that intentionally shadow standard R6RS or Chez Scheme functions with the same name:
  (get-bytevector-all) (get-bytevector-n) (get-bytevector-some) (get-char) (get-datum) (get-line) (get-string-all) (get-string-n) (get-u8) (put-bytevector) (put-bytevector-some) (put-char) (put-datum) (put-string) (put-string-some) (put-u8)
  They are compatible replacements, and add the ability to omit the port argument, which defaults to (current-input-port) for (get-...) functions and to (current-output-port) for (put-...) functions
• add functions (getenv) (putenv) that intentionally shadow standard R6RS or Chez Scheme functions with the same name:
  They are intended as replacements, and access environment variables of job-id #f i.e. either (sh-current-job) if it's set, or as fallback (sh-globals) that represents the main schemesh process
• allow getting/setting port position in ports returned by functions (open-fd-redir-...-port) (open-file-binary-input-port) (open-file-utf8b-input-port) (current-input-port) (current-out-port) (current-error-port) (sh-stdin) (sh-stdout) (sh-stderr) if the underlying file descriptor supports seeking
• improve the following functions to also accepts bytespan paths in addition to bytevector, string and charspan:
  (delete-directory) (delete-file) (directory-list) (directory-list-type) (mkdir) (file-delete) (file-directory?) (file-exists?) (file-rename) (file-regular?) (file-type) (open-file-fd) (open-file-utf8b-input-port) (text->bytevector) (text->bytevector0)
• improve source code conformance to R6RS, and start creating an incomplete Racket compatibility layer to use when compiling schemesh from Racket #!r6rs mode

Fixes:

• fix issue #12: a bug in shell syntax when parsing assignment after backslash+newline
• fix macro (in-hash-pairs)

New features:

• replace standard Scheme textual ports (current-input-port) (current-output-port) (current-error-port)
  with interruptible ones that honor job redirections and use UTF-8b
• export several new functions to iterate on lists, hashtables, strings, etc.
• add script utils/show_system_info.sh and mention it in README.md when opening an issue

Improvements:

• fix job control on Scheme jobs $() running inside a pipeline
• update Scheme binary ports (sh-stdin) (sh-stdout) (sh-stderr) to be interruptible and honor job redirections
• internally rewrite and cleanup POSIX signal handlers

New features:

• add function (sh-expr) to create shell jobs that execute arbitrary Scheme code, and the corresponding macro (shell-expr)
• implement job control also on arbitrary Scheme code running inside a (shell-expr)
• update both shell parser and Scheme parser to expand $( ) to (shell-expr ( ))

Improvements:

• ignore newlines at the beginning of shell syntax: allows writing Scheme code in a new line under the REPL prompt
• ignore newlines after shell syntax tokens { [ ! ; & && || | |& < <> > >> <& >& - in most cases, this removes the need to end a line with \ for continuing a shell command in the next line
• always queue job status change notifications for later displaying them, instead of sometimes displaying them immediately

Removed features:

• remove function (sh-resume) and subsume it into (sh-wait)
• rename shell builtin expr -> value
• remove function (sh-bool)

Fixes:

• fix (charhistory-save-to-path) not to raise exceptions: it was preventing schemesh
  from exiting if ~/.cache/schemesh/ directory does not exist or is not writable