mirror of
https://github.com/wader/fq.git
synced 2024-11-29 23:27:12 +03:00
e3ae1440c9
Feels less cluttered, easier to read and more consistent. Still keep tovalue, tobytes etc that are more basic functions this only renamed format related functions. Also there is an exceptin for to/fromjson as it comes from jq. Also fixes lots of spelling errors while reading thru.
85 lines
2.5 KiB
Plaintext
85 lines
2.5 KiB
Plaintext
def protobuf_to_value:
|
|
.fields | map({(.name | tostring): (.enum // .value)}) | add;
|
|
|
|
# converted from https://github.com/FFmpeg/FFmpeg/blob/870bfe16a12bf09dca3a4ae27ef6f81a2de80c40/libavutil/display.c av_display_rotation_get
|
|
def mp4_matrix_structure_rotation:
|
|
( .a as $s0
|
|
| .c as $s3
|
|
| .b as $s1
|
|
| .d as $s4
|
|
| ($s0*$s0 + $s3*$s3 | sqrt) as $scale0
|
|
| ($s1*$s1 + $s4*$s4 | sqrt) as $scale1
|
|
| atan2($s1/$scale1; $s0 / $scale0) * 180 / 3.14159265359
|
|
| -round
|
|
);
|
|
|
|
def flac_dump:
|
|
[ "fLaC"
|
|
, first(.. | select(format == "flac_metadatablocks"))
|
|
, (.. | select(format == "flac_frame"))
|
|
] | tobits;
|
|
|
|
def urldecode:
|
|
gsub(
|
|
"%(?<c>[a-fA-F0-9]{2})";
|
|
( .c
|
|
| ascii_downcase
|
|
| explode
|
|
# "0"-"9" or "a"-"f"
|
|
| map(.-48 | if .>=49 then .-39 else . end)
|
|
| [.[0]*16+.[1]]
|
|
| implode
|
|
)
|
|
);
|
|
|
|
# see all first frames where sample_rate changes
|
|
# ex: .frames | changes(.header.sample_rate)
|
|
def changes(f): streaks_by(f)[].[0];
|
|
|
|
def to_radix62sp: to_radix(62; "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ");
|
|
def from_radix62sp: from_radix(62; {
|
|
"0": 0, "1": 1, "2": 2, "3": 3,"4": 4, "5": 5, "6": 6, "7": 7, "8": 8, "9": 9,
|
|
"a": 10, "b": 11, "c": 12, "d": 13, "e": 14, "f": 15, "g": 16,
|
|
"h": 17, "i": 18, "j": 19, "k": 20, "l": 21, "m": 22, "n": 23,
|
|
"o": 24, "p": 25, "q": 26, "r": 27, "s": 28, "t": 29, "u": 30,
|
|
"v": 31, "w": 32, "x": 33, "y": 34, "z": 35,
|
|
"A": 36, "B": 37, "C": 38, "D": 39, "E": 40, "F": 41, "G": 42,
|
|
"H": 43, "I": 44, "J": 45, "K": 46, "L": 47, "M": 48, "N": 49,
|
|
"O": 50, "P": 51, "Q": 52, "R": 53, "S": 54, "T": 55, "U": 56,
|
|
"V": 57, "W": 58, "X": 59, "Y": 60, "Z": 61
|
|
});
|
|
|
|
# "01:09:55.76" -> 4195.76
|
|
# 4195.76 -> "01:09:55.76"
|
|
def duration:
|
|
def lpad($s; $w): ($s * ($w+1-length))[1:] + .;
|
|
def _string:
|
|
( split(":")
|
|
| map(tonumber)
|
|
| reverse
|
|
| [foreach .[] as $n (0; . + 1; pow(60; .-1) * $n)]
|
|
# sum smallest to largest seem to improve precision
|
|
| sort
|
|
| reverse
|
|
| add
|
|
);
|
|
def _number:
|
|
if . == 0 then 0
|
|
else
|
|
[ ( [ (recurse(if . > 0 then intdiv(.; 60) else empty end) | . % 60)]
|
|
| reverse
|
|
| .[1:]
|
|
| map(tostring | lpad("0"; 2))
|
|
| join(":")
|
|
)
|
|
# ugly but float is not accurate enough
|
|
, ( tostring
|
|
| split(".")[1] // empty
|
|
| ".", .
|
|
)
|
|
] | join("")
|
|
end;
|
|
if type == "string" then _string
|
|
elif type == "number" then _number
|
|
else error("expected string or number") end;
|