🛠 A Guy of Many Trades 🛠
Table of Contents generated with DocToc
🛠 A Guy of Many Trades 🛠
Structure
- Only Peer-Dependencies (except
cnd
, intertype
) - Sub-libraries accessible as
GUY.${library_name}
- Most sub-libraries implemented using
GUY.props.def_oneoff()
, therefore dependencies (which are declared
peer dependencies) will only be require()
d when needed.
Modules
GUY.props
: Common Operations on Object Properties
-
GUY.props.def: ( target, name, cfg ) ->
is just another name for Object.defineProperty()
.
-
GUY.props.hide: ( object, name, value ) ->
is a shortcut to define a non-enumerable property as in
Object.defineProperty object, name, { enumerable: false, value, }
.
-
GUY.props.xray: ( owner, host ) ->
—A little bit the inverse to GUY.props.hide()
, it will return an
object with all key/value pairs set that were found on owner
after applying GUY.props.keys owner, { hidden: true, symbols: true, builtins: false, }
. When host
is set, that value will be used to store the
key/value pairs in, possibly overwriting existing keys. This method is useful for printing and debugging
objects with non-enumerable and/or symbol keys.
-
GUY.props.def_oneoff: ()
-
GUY.props.pick_with_fallback = ( d, fallback, keys... ) ->
—Given an object d
, a fallback
value and
some keys
, return an object that whose keys
are the ones passed in, and whose values are either the
same as found in d
, or fallback
in case a key is missing in d
or set to undefined
. If d[ key ]
is null
, it will be replaced by fallback
. When no keys are given, an empty object will be returned.
-
GUY.props.nullify_undefined = ( d ) ->
—Given an object d
, return a copy of it where all undefined
values are replaced with null
. In case d
is null
or undefined
, an empty object will be returned.
-
GUY.props.omit_nullish = ( d ) ->
—Given an object d
, return a copy of it where all undefined
and
null
values are not set. In case d
is null
or undefined
, an empty object will be returned.
-
GUY.props.crossmerge = ( cfg ) ->
—Given an object keys
and an object values
, return a new object
whose keys come from the former and whose values come fom the latter. Should there be a key in keys
that
is not set in values
, an error will be thrown unless fallback
has been set (to any value, inlcuding
undefined
). This method calls GUY.props.keys()
and will accept the cfg
settings used there
(allow_any
, symbols
, builtins
, hidden
, depth
).
-
GUY.props.has = ( x, key ) ->
—Given any value x
, return whether the value has a given property
(key
). This is a safe version of Reflect.has()
that never throws an error. Like direct property access
(using x.key
or x[ 'key' ]
) but unlike Object.getOwnPropertyDescriptor()
&c, GUY.props.has()
looks
into the prototype chain; like Object.getOwnPropertyDescriptor()
, it does not trigger property getters.
-
GUY.props.get = ( x, key, fallback ) ->
—Given any value x
, return the value of property named in
key
. If that property is missing, throw an error, but when fallback
has been given, return fallback
instead. Using GUY.props.get x, 'foo'
is like saying x.foo
or x[ 'foo' ]
except that it doesn't
tolerate missing property keys; using it with a fallback as in GUY.props.get x, 'foo', undefined
is like
saying x.foo
or x[ 'foo' ]
except that it also works for null
and undefined
.
Using GUY.props.has()
and GUY.props.get()
it is always possible to circumvent errors being thrown and
instead do value-based error handling (and raise one own's errors where seen fit). One pattern to do so is
to define a private Symbol instead of relying on undefined
that could have been caused by all kinds of
circumstances. Here's a real-life example; it was created in a context where instances of the Strict_owner
class (for which see below) are prevalent (i.e. those objects will throw an error instead of returning
undefined
as is the standard in JS). This function will return the length
of a value x
where that
attribute is present (as in strings and arrays), or else its size
where that is present (as in sets and
maps), else it will return fallback
where given, or resort to throwing an error:
notavalue = Symbol 'notavalue'
size_of = ( x, fallback = misfit ) ->
return R unless ( R = GUY.props.get x, 'length', notavalue ) is notavalue
return R unless ( R = GUY.props.get x, 'size', notavalue ) is notavalue
return fallback unless fallback is misfit
throw new Error "expected an object with `x.length` or `x.size`, got a #{typeof x} with neither"
GUY.props.resolve_property_chain = ( owner, property_chain ) ->
—Given an owner
value (typically an
object) and a property_chain
as a list of names, repeatedly apply the names, first against the owner,
then the next name against that result and so on. This allows one to programmatically resolve chained
property access and to parametrize a literal owner.first.second.other.last
as
GUY.props.resolve_property_chain owner, [ 'first', 'second', 'other', 'last', ]
.
GUY.props.keys()
GUY.props.keys: () =>
—Like Object.keys( t )
, Reflect.ownKeys( t )
(which is equivalent to
Object.getOwnPropertyNames( target ).concat(Object.getOwnPropertySymbols( target ))
) but more versatileGUY.props.keys()
can retrieve or skip non-enumerable (a.k.a. 'hidden') keysGUY.props.keys()
can retrieve only own keys (with { depth: 0 }
) or descend into the prototype chain
for any number of steps: { depth: null, }
(the default, equivalent to { depth: Infinity, }
) will
descend into the prototype chain or e.g. { depth: 1, }
GUY.props.keys()
can retrieve string keys as well as symbol keys with { symbols: true, }
GUY.props.keys()
works for all JS values, including null
and undefined
when { allow_any: true, }
is set (the default)GUY.props.keys()
can retrieve hidden (i.e. non-enumerable) keys with { hidden: true, }
GUY.props.keys()
can retrieve builtins with { builtins: true, }
; observe that since builtins are
always hidden, { builtins: true, hidden: false, }
makes no sense and causes a validation error.
lst = [ 'x', ]
( Object.keys lst )
# [ '0', ]
( ( k for k of lst ) )
# [ '0', ]
( GUY.props.keys lst )
# [ '0', ]
( GUY.props.keys lst, { symbols: true, builtins: true, } )
# [ '0', 'length', 'constructor', 'concat', 'copyWithin', 'fill', 'find', 'findIndex', 'lastIndexOf',
# 'pop', 'push', 'reverse', 'shift', 'unshift', 'slice', 'sort', 'splice', 'includes', 'indexOf', 'join',
# 'keys', 'entries', 'values', 'forEach', 'filter', 'flat', 'flatMap', 'map', 'every', 'some', 'reduce',
# 'reduceRight', 'toLocaleString', 'toString', 'at', 'findLast', 'findLastIndex', Symbol.iterator,
# Symbol.unscopables, '__defineGetter__', '__defineSetter__', 'hasOwnProperty', '__lookupGetter__',
# '__lookupSetter__', 'isPrototypeOf', 'propertyIsEnumerable', 'valueOf', '__proto__' ]
( GUY.props.keys Array, { symbols: true, builtins: true, } )
# [ 'length', 'name', 'prototype', 'isArray', 'from', 'of', Symbol.species, 'arguments', 'caller',
# 'constructor', 'apply', 'bind', 'call', 'toString', Symbol.hasInstance, '__defineGetter__',
# '__defineSetter__', 'hasOwnProperty', '__lookupGetter__', '__lookupSetter__', 'isPrototypeOf',
# 'propertyIsEnumerable', 'valueOf', '__proto__', 'toLocaleString' ]
GUY.props.tree()
GUY.props.tree()
is the logical dual to GUY.props.keys()
: instead of descending into the prototype
chain, tree()
climbs through the key/value pairs attached to an object and, recursively, through the trees
of each value's key/value pairs. In default mode, the method returns a list of lists of names ('paths'):
log = console.log
{ inspect, } = require 'util'
d = { a: [ 0, 1, 2, ], e: { g: { some: 'thing', }, h: 42, h: null, }, empty: {}, }
for path in GUY.props.tree d
log inspect path
This will print:
[ 'a' ]
[ 'a', '0' ]
[ 'a', '1' ]
[ 'a', '2' ]
[ 'e' ]
[ 'e', 'g' ]
[ 'e', 'g', 'some' ]
[ 'e', 'h' ]
[ 'empty' ]
The default configuration for the tree()
method is { allow_any: true, symbols: false, builtins: false, hidden: false, depth: null, evaluate: null, sep: null, }
, with the last two settings being specific to
tree()
, and the first five having the same meaning as for GUY.props.keys()
, q.v.
To turn the result into a list of strings, pass in a property sep
:
...
for path in GUY.props.tree d, { sep: '.', }
...
'a'
'a.0'
'a.1'
'a.2'
'e'
'e.g'
'e.g.some'
'e.h'
'empty'
The evaluate
setting can be set to a function which will be called with an object { owner, key, value, }
where owner
is the current object being iterated over, and key
/value
the current property name and
value. The evaluate function should return true
, false
, or a string that may contain the words take
and/or descend
; the default (when no evaluate
function is given) is take,descend
, meaning all keys
will be both included in the list as well as descended into. true
is equivalent to take,descend
; false
means skip this property altogether. In the below example, we use the evaluate
setting to avoid descending
into arrays and to include only the tips (leaves, endpoints) of the tree:
evaluate = ({ owner, key, value, }) ->
return 'take' if Array.isArray value
return 'take' unless GUY.props.has_keys value
return 'descend'
...
for path in GUY.props.tree d, { evaluate, sep: '.', }
...
'a'
'e.g.some'
'e.h'
'empty'
Class for Strict Ownership
JavaScript is famously forgiving when it comes to accessing non-existing object properties. However this
lenience is also conducive to silent failure. Strict_owner
is an ES6 class that aims to provide users with
a convenient mechanism to produce object that throw an error when a non-existing property is being accessed.
When you extend your class with GUY.props.Strict_owner
, instance of your class will now throw an error
when a non-existing property is accessed.
Note Most often one will want to define a class that extends Strict_owner
, however, it is also
possible to pass in an arbitrary object—including a function—as property target
to the constructor, e.g.
f = ( x ) -> x * 2
x = new GUY.props.Strict_owner { target: f, }
Note As of Guy v6, special methods get()
and set()
have been removed from Strict_owner
. They
have been replaced with a much cleaner and more correct implementation as GUY.props.get()
and
GUY.props.has()
. These methods are more versatile, too, since they can be used with any JS value
(including the always-problematic null
and undefined
).
Note GUY.props.Strict_owner
is implemented using an ES6
Proxy
. While
proxies are powerfool tools they're also notoriously hard to get
right because of the many methods one could/should/must
implement and which all have to act in mutually consistent ways. As a result, my personal recommendation
here is to try hard not use a proxy and use another methodology to achieve a workable solution where
possible. One thing that will probably not work is proxying a proxy: When you have a Strict_owner
object
(which already is a proxy) and wrap it in another proxy, chances are you'll get errors like instance does not have property '0'
when attempting to use node:util.inspect()
on that instance. I have been so far
unable to fix that bug and since been looking for a solution that works without the need for a double
proxy in that other codebase.
In order to get an object which is sealed and or frozen, use properties seal
and freeze
, respectively:
d = { x: 42, }
dso = new GUY.props.Strict_owner { target: d, seal: true, freeze: true, }
# this works:
dso.x # 42
# but none of these work:
# Strict owner ship means unknown properties cannot be accessed:
dso.y # Strict_owner instance does not have property 'y'
# Properties of a frozen object cannot be re-assigned:
dso.x = 48 # TypeError: Cannot assign to read only property 'x'
# Properties cannot be added to a sealed object (also implied by being frozen):
dso.y = 'something' # TypeError: Cannot define property y, object is not extensible
In addition, when property oneshot
is set (and neither seal
nor freeze
), properties can be set once,
but not get re-assigned. In case non-re-assignable values should also be immutable, consider to set frozen
objects:
d = { x: 42, }
dso = new GUY.props.Strict_owner { target: d, oneshot: true, }
dso.xy = new GUY.props.Strict_owner { target: { foo: 'bar', }, freeze: true, }
# dso.x = 123 # Strict_owner instance already has property 'x'
# dso.xy = {} # Strict_owner instance already has property 'xy'
# dso.xy.foo = 'gnu' # TypeError: Cannot assign to read only property 'foo'
Observe that because all of the sealing, freezing and one-shot business is performed on the proxy, not on
the target object, we can still manipulate that one:
debug '^35345^', dso # { x: 42, xy: { foo: 'bar' } }
debug '^35345^', d # { x: 42, xy: { foo: 'bar' } }
# we *can* still manipulate the underlying object:
d.x = 123
debug '^35345^', dso # { x: 123, xy: { foo: 'bar' } }
debug '^35345^', d # { x: 123, xy: { foo: 'bar' } }
It is possible to lift the strict behavior of Strict_owner
instances by using Strict_owner.set_locked false
and resume strict behavior with Strict_owner.set_locked true
.
Special Keys
return "#{instance.constructor.name}" if key is Symbol.toStringTag
return target.constructor if key is 'constructor'
return target.toString if key is 'toString'
return target.call if key is 'call'
return target.apply if key is 'apply'
return target[ Symbol.iterator ] if key is Symbol.iterator
return target[ node_inspect ] if key is node_inspect
return target[ 0 ] if key is '0'
GUY.async
: Asynchronous Helpers
These 'five letter' methods are convenience methods in the sense that they are very thin shims over the
somewhat less convenient JavaScript methods. For many people, the most strightforward way to understand what
these methods do will be to read the very simple definitions:
every: ( dts, f ) -> setInterval f, dts * 1000
after: ( dts, f ) -> setTimeout f, dts * 1000
cease: ( toutid ) -> clearTimeout toutid
sleep: ( dts ) -> new Promise ( done ) => setTimeout done, dts * 1000
defer: ( f = -> ) -> await sleep 0; return await f()
In each case, dts
denotes an interval (delta time) measured in seconds (not milliseconds) and f
denotes a function. every()
and after()
return so-called timeout IDs (toutid
s), i.e. values that are
recognized by cease()
(clearTimeout()
, clearInterval()
) to stop a one-off or repetetive timed function
call. sleep()
returns a promise that should be awaited as in await sleep 3
, which will allow another
task on the event loop to return and resume execution no sooner than after 3000 milliseconds have elapsed.
Finally, there is defer()
, which should also be await
ed. It is a special use-case of sleep()
where the
timeout is set to zero, so the remaining effect is that other tasks on the event loop get a chance to run.
It accepts an optional function argument whose (synchronous or asynchronous) result will be returned.
GUY.nowait
: De-Asyncify JS Async Functions
Note Due to ongoing issues when compiling the deasync
module that this functionality
is implemented in, guy-nowait
has been removed from this release.
**Peer Dependencies**: [`abbr/deasync`](https://github.com/abbr/deasync)
-
GUY.nowait.for_callbackable: ( fn_with_callback ) ->
—given an asynchronous function afc
that
accepts a NodeJS-style callback (as in afc v1, v2, ..., ( error, result ) -> ...
), returns a synchronous
function sf
that can be used without a callback (as in result = sf v1, v2, ...
).
-
GUY.nowait.for_awaitable: ( fn_with_promise ) ->
—given an asynchronous function afp
that can be
used with await
(as in result = await afp v1, v2, ...
) returns a synchronous function f
that can be
used without await
(as in result = sf v1, v2, ...
).
GUY.process
: Process-Related Utilities
Peer Dependencies: sindresorhus/exit-hook
GUY.process.on_exit: ( fn ) => ...
—call fn()
before process exits. Convenience link for
sindresorhus/exit-hook
, which see for details. Note
When installing this peer dependency, make sure to do so with the last CommonJS version added, as in npm install exit-hook@2.2.1
.
GUY.lft
: Freezing Objects
GUY.left.freeze()
and GUY.lft.lets()
provide access to the epynomous methods in
letsfreezethat
. freeze()
is basically
Object.freeze()
for nested objects, while d = lets d, ( d ) -> mutate d
provides a handy way to mutate
and re-assign a copy of a frozen object. See the
documentation for details.
GUY.fs
: File-Related Stuff
-
GUY.fs.walk_lines = ( path, cfg ) ->
—Given a path
, return a synchronous iterator over file
lines. This is the most hassle-free approach to synchronously obtain lines of text files in NodeJS that
I'm aware of, yet. The optional cfg
argument may be an object with a single property decode
; when set
to false
, walk_lines()
will iterate over buffers instead of strings.
-
GUY.fs.walk_circular_lines = ( path, cfg ) ->
—Given a path
, return an iterator over the lines in
the referenced file; optionally, when the iterator is exhausted (all lines have been read), restart from
the beginning. cfg
may be an object with the keys:
loop_count
—(cardinal; default: 1
) controls how many times to loop over the file. Set to
+Infinity
to allow for an unlimited number of laps.line_count
—(cardinal; default: +Infinity
) controls the maximum number of lines that will be
yielded.- The iteration will finish as soon as the one or the other limit has been reached.
- By default,
GUY.fs.walk_circular_lines()
will act like GUY.fs.walk_lines
. - The iterator will not yield anything when either
loop_count
or line_count
are set to 0
.
-
GUY.fs.get_content_hash = ( path, cfg ) ->
—Given a path
, return the
hexadecimal sha1
hash digest for its contents. On Linux, this uses sha1sum
, and shasum
on all
other systems.
GUY.src
: JS Source Code Analysis
This submodule needs peer-dependencies, install them with
npm install acorn acorn-loose acorn-walk astring
or
pnpm add acorn acorn-loose acorn-walk astring
-
@STRICT_PARSER = require 'acorn'
-
@LOOSE_PARSER = require 'acorn-loose'
-
@AST_walk = require 'acorn-walk'
-
@ASTRING = require 'astring'
-
@parse = ( cfg ) =>
—Given either a JS source text
or a function
, return an
ESTree-compliant AST. Should an error occur and fallback
is set to
any value, that value will be returned; otherwise, the error will be thrown. The use
parameter controls
which parser is used and can take on the values 'strict'
, 'strict,loose'
, and 'loose'
. For many
settings 'strict,loose'
will probably the right setting since the strict parser will balk already on
unnamed function declarations that are not part of assignment, while the 'loose' parser happily (and
correctly parses those). For this reason, the default setting is use: 'strict,loose'
.
-
slug_node_from_simple_function = ( cfg ) =>
—Same as slug_from_simple_function()
, below, but
returns an AST node representing the result. You can manipulate the node if you want it and then pass it
to GUY.src._generate()
to get it rendered as JS, but the result will slightly differ from what
slug_from_simple_function()
would return for the same input because that function does some
post-processing on the source text to make it even terser.
-
slug_from_simple_function = ( cfg ) =>
—Given the same cfg
object one would use for
GUY.src.parse()
, return a 'slug' (i.e. a condensed form) of its source text. This slug is defined to be
- if function has no return:
undefined
- if function has single
return
: argument
property of the ReturnStatement
node - if function has several
return
s: firstlast BlockStatement
(i.e. the function
body) (???)
Examples:
█ ( -> )
''
█ ( ( x ) -> 42 )
'42'
█ ( ( x ) -> ( not x? ) or ( @isa.object x ) or ( @isa.nonempty.text x ) )
'x == null || this.isa.object(x) || this.isa.nonempty.text(x)'
█ ( `function ( x ) { 42; }` )
'42;'
█ ( `function ( x ) { return 42; }` )
'42'
█ ( ( x ) -> if x? then true else false )
'if (x != null) { return true; } else { return false; }'
█ ( ( x ) -> ( not x? ) or ( @isa.object x ) or ( @isa.nonempty.text x ) )
'x == null || this.isa.object(x) || this.isa.nonempty.text(x)'
█ f3
'if (x > 0) { return true; } if (x < 0) { return false; } return null;'
GUY.trm
- Preview version, expect changes
- colorize terminal output
- two variants of
node:util.inspect()
: GUY.trm.inspect()
, GUY.trm.rpr()
- standardized loggers:
alert()
, debug()
, help()
, info()
, plain()
, praise()
, urge()
, warn()
,
whisper()
- produce with
GUY.trm.get_loggers badge
where badge
identifies your submodule - includes short timestamp
- writer
GUY.trm.log()
writes undecorated stuff to process.stderr()
- writer
GUY.trm.echo()
writes undecorated stuff to process.stdout()
- all writers and loggers apply
GUY.trm.rpr
to each argument independently, so echo a, b, c
is like
console.log ( rpr a ), ( rpr b ), ( rpr c )
- writer
GUY.trm.pen()
returns string representation as used in GUY.trm.log()
&c GUY.trm.strip_ansi: ( text ) ->
uses RegEx from
(chalk)[https://raw.githubusercontent.com/chalk/ansi-regex/main/index.js] to strip ANSI codes from a given
string
GUY.sets
Operations on sets, copied / modelled on JavaScript for impatient programmers (ES2022 edition): Missing
Set operations
unite: ( P... ) ->
—return the union of all sets passed inintersect: ( P... ) ->
—return the intersection of all sets passed insubtract: ( a, b ) ->
—return the set of elements of set a
except for those that are also in set
b
To Do
-
[–] adopt icql-dba/errors#Dba_error
:
class @Dba_error extends Error
constructor: ( ref, message ) ->
super()
@message = "#{ref} (#{@constructor.name}) #{message}"
@ref = ref
return undefined
-
[–] while test @[ "nowait with async steampipes" ]
works in isolation, running the test suite hangs
indefinitely.
-
[+] see whether n-readlines
may be replaced by a simpler, faster implementation as it does
contain some infelicitous loops in
_searchInBuffer
that can
probably be replaced by
buffer.indexOf()
.
Also, there's a similar implementation in
intertext-splitlines
.
Benchmarks
show that patched version with suitable chunk size performs OK; using patched version to avoid deprecation
warning.
-
[–] GUY.fs.walk_lines()
: allow to configure; make trimEnd()
the default
-
implement easy way to collect, rediect process.stdout
, process.stderr
:
original_stderr_write = process.stderr.write.bind process.stderr
collector = []
process.stderr.write = ( x ) ->
echo '^3453^', type_of x
# FS.writeSync output_fd, x
collector.push x
original_stderr_write '^1234^ '
original_stderr_write x
echo "(echo) helo world"
info "(info) helo world"
info "whatever goes on here"
warn CND.reverse "is collected"
for x in collector
process.stdout.write '^collector@4565^' + x
-
[–] take over tabulation (as in hengist/src/helpers
)
-
[–] could the SQL
string annotation / tagged literal function be syntactically extended to allow
simpler interpolation of escaped names? Could we instantiate it with a dictionary of values?
-
[–] GUY.src.get_first_return_clause_text()
:
- [–] change input format to standard
cfg
-based to make compatible with call conventions for
GUY.src.parse()
- [–] use
fallback
argument to decide whether to return value or throw error in case of parsing
failure (same for parse()
) - [–] return based on how many
ReturnStatement
s are found:
- if function has no return:
undefined
- if function has single
return
: argument
property of the ReturnStatement
node - if function has several
return
s: firstlast BlockStatement
(i.e. the function
body) (???)
-
[–] move GUY.src._generate()
to public API
-
[–] what should be the correct output for GUY.src.slug_from_simple_function { function: ( -> ), }
?
Currently it is the empty string which is not ideal
-
[–] use
Reflect.has()
to check for property availability for Strict_owner
s instead of using instance method; provide as
GUY.props.has()
-
[–] likewise, use GUY.props.get: ( target, name, fallback = misfit ) ->
instead of instance method
-
[–] consider to move submodule _builtins
, test for builtins to Intertype, backport to
Intertype-legacy
-
[–] consider to make trm.log()
write to stdout
, err()
to stderr
-
[–] implement trm.write()
write to stdout
without trailing newline (but formatting like log()
)
-
[–] consider to drop trm.pen()
, use improved GUY.trm.rpr()
instead
-
[–] fix GUY.props.get '', 'length'
(works for sets but not for string)s
-
[–] consider to re-implement deep_copy()
from letsfreezethat
using
-
[–] remove dependencies as far as possible, make GUY run in browser
LOUPE = require '../deps/loupe.js'
@rpr = rpr = ( x ) => LOUPE.inspect x, { customInspect: false, }
@xrpr = ( x ) -> ( rpr x )[ .. 1024 ]
-
[–] modify behavior of GUY.props.tree()
:
- [–] callbacks are
take_key()
, descend_value()
, filter_path()
- [–]
take_key: ( owner, key, value ) ->
- [–]
descend_value: ( owner, key, value ) ->
- [–]
filter_path: ( owner, path ) ->
; path will be list of strings w/out sep
, single string with it - [+] implement
GUY.props.walk_tree()
-
[–] modify behavior of GUY.trm.rpr()
:
- [–] colorize for readablity
- [–] change signature to either
rpr x, cfg
or rpr P...
- [–] in either case, provide a way to pass configuration to
node:util.inspect
- [–] allow indented output
-
[–] implement method that allows to name a type and give a cfg object, returns cfg for named typed
based on crossmatch()
ing defaults for that type with given cfg
-
[–] GUY.trm.warn()
, alert()
: collect and re-issue on process exit; may want to throw error or at
least issue non-zero error code when messages beyond (configurable) urgency threshold were issued; this
will allow apps to not bail out prematurely on minor issues and keep messages from getting hidden in
regular messages
-
[–] turn Strict_owner
instances into sealed objects, retaining old behavior in new class
Strict_getter
-
[–] turn GUY
, submodules into Strict_owner
instances
-
[–] move deep freezing, deep sealing from lft
to props
?
-
[–] concerning Strict_owner
: in
hengist/dev/intertype/_ng.test.coffee
we have found a way to build custom-named functions with strict ownership:
@_demo_type_cfgs_as_funmctions_2 = ->
class Intertype
create_type_cfg: ( cfg ) ->
defaults = { extras: true, collection: false, type: null, }
cfg = { defaults..., cfg..., }
name = cfg.type
R = ( ( x ) -> x ** 2 ).bind @
Object.defineProperty R, 'name', { value: name, }
R[ k ] = v for k, v of cfg
R = new GUY.props.Strict_owner target: R
Object.seal R
# R = GUY.lft.freeze R1 = R # <== doesn't freeze???
Object.freeze R # <== works as expected
return R
types = new Intertype()
urge '^982-1^', f = types.create_type_cfg { type: 'foobar', }
# Object is frozen, sealed, and has a strict `get()`ter:
urge '^982-2^', Object.isFrozen f
urge '^982-3^', Object.isSealed f
try f.collection = true catch error then warn rvr error.message # Cannot assign to read only property 'collection' of function 'function () { [native code] }'
try f.xxx catch error then warn rvr error.message # ^guy.props.Strict_owner@1^ Strict_owner instance does not have property 'xxx'
try f.xxx = 111 catch error then warn rvr error.message # Cannot define property xxx, object is not extensible
info '^982-4^', f.name
info '^982-5^', f 42
return null
This should lead to the following:
- [–] fix the apparent bug that
GUY.lft.freeze()
does not freeze a function - [–] an extension of
GUY.props.Strict_owner
:
- [+] incorporate (deep?)
seal
ing (next to freeze
ing) - [–] allow to pass in a 'deep target' such that properties not found on the immediate target (the
named function in the above) will also be searched in the deep target (such as the
Intertype
or
Dbay
instance), mimicking a prototype chain
Is Done
- [+] make choice between parsers configurable:
- only
acorn
- first
acorn
, upon parse error acorn-loose
- only
acorn-loose
- [+]
parse()
: use fallback
argument to decide whether to return value or throw error in case of
parsing failure - [+] testing for a key without inadvertantly retrieving its value is surprisingly involved in JS.
Re-implement
GUY.props.has()
- (1) using
Reflect.has()
catching TypeError: Reflect.has called on non-object
, and
Object.getPrototypeOf()
, or - (2) using attribute access
x[ key ]
, discarding value and catching errors from strict owners, null
and undefined
. - option (2) should be faster, maybe just live with the fact that attribute checking without value
retrieval is not very viable in JS
- depending on favorable benchmarks, may want to cache whether instances of a given type are collections
with a size and if so, which property (
length
or size
) is used
structuredClone
; benchmarks
- [+] rename
GUY.props.has_keys()
-> GUY.props.has_any_keys()