View binary array data stored in files, from node, the browser and the command line.
npm install -g arrayviewer
Show the 4th element (-i 3) in an array stored in a file at ./data/a.f32:
arrayviewer ./data/a.f32 -i 3
produces something like,
[126.48208618164062, 127.23143005371094, 136.79074096679688,
-->126.48942565917969
127.26338195800781, 136.84552001953125, 126.47312927246094, 127.27062225341797, ...]
Show the 10th element, with more context (-c 5).
arrayviewer ./data/a.f32 -i 9 -c 5
might produce,
[..., 136.84552001953125, 126.47312927246094, 127.27062225341797, 136.86407470703125, 126.44374084472656,
-->127.25633239746094
136.88494873046875, 126.36851501464844, 127.19410705566406, 136.8431396484375, 126.31245422363281, ...]
Show some extra information with -v,
arrayviewer ./data/a.f32 -i 9 -c 5 -v
Length: 150528
[..., 136.84552001953125, 126.47312927246094, 127.27062225341797, 136.86407470703125, 126.44374084472656,
-->127.25633239746094
136.88494873046875, 126.36851501464844, 127.19410705566406, 136.8431396484375, 126.31245422363281, ...]
Array type is inferred from a file's extension and can be overridden with the -t
option.
arrayviewer ./data/a.arr -t int32
Extensions are mapped to a TypedArray in the
following way,
| extension | TypedArray | type |
|---|---|---|
| i8 | Int8Array |
int8 |
| u8 | Uint8Array |
uint8 |
| i16 | Int16Array |
int16 |
| u16 | Uint16Array |
uint16 |
| i32 | Int32Array |
int32 |
| u32 | Uint32Array |
uint32 |
| f32 | Float32Array |
float32 |
| f64 | Float64Array |
float64 |
or (for non-binary types) to Javascript types like this,
| extension | Result Type | type |
|---|---|---|
| json | Object | json |
| key | Object | json |
| txt | String | str |
| csv | String | str |
| tsv | String | str |
Any value from the type column may be supplied with the -t option.
If none of these match the file extension (and no explicit type or metadata file is provided), the data will be interpreted as a Uint8Array.
arrayviewer ./data/a.f32 -i 9 -c 6 -m
If the -m option is specified it will also look for a file of the same name
as the array with the .meta extension. The meta file has the following format
{
"shape" : [224, 224, 3],
"type" : "float32"
}type should be a string containing any of the values listed in the "type" column from the tables above.
These values match the numpy dtypes
In addition to allowing you to specify a type,
providing a meta file allows you to index into an array using
-i, -j, -k to specify row, column and channel, respectively.
Write compatible arrays from numpy like this,
# given array 'a'
f = open('./data/a.f32', 'wb')
f.write(a.astype(np.float32).tostring())
f.close