Ignore bits that were introduced as padding in transformUnitSize (f…

…ixes wierkstudio#40)

Base64 embeds bytes into 6 bit units. It takes two units to properly encode a single byte. If you then decode this, you can conclude that the 4 additional bits contain no information. (Unless the input violates our assumption that it is a base64 encoding.)

You can conclude that if $y = \lceil xs_{input} / s_{output}\rceil$ with x an integer, then $x <= \lfloor ys_{output} / s_{input}\rfloor$. This commit checks every added unit if it is the last, and doesn't add the null byte if it is. It might be more efficient to just use the larger array and slice off the null byte when it exists.

It can be impossible to determine a true zero byte from padding if `inputSize < outputSize`, but for binary-to-text that requires an alphabet of more than 256 characters.

Signed-off-by: anderium <[email protected]>

extensions/essentials/src/binary-slice.ts

@@ -68,15 +68,15 @@ const execute: OperationExecuteExport = (request) => {
 
       // Transform the array of bytes to an array of bits and slice them
       // precisely to create the bit slice
-      const bitSlice = transformUnitSize(byteSlice, 8, 1).slice(
+      const bitSlice = transformUnitSize(byteSlice, 8, 1, true).slice(
         bitSliceArgs.start - byteSliceStart * 8,
         bitSliceArgs.end - byteSliceStart * 8
       )
 
       // Fill up the first byte with zero bits and turn the bits back into bytes
       const paddingBits = bitSlice.length % 8 > 0 ? 8 - bitSlice.length % 8 : 0
       const sliceBitsPadded = new Array(paddingBits).fill(0).concat(bitSlice)
-      const sliceBytes = transformUnitSize(sliceBitsPadded, 1, 8)
+      const sliceBytes = transformUnitSize(sliceBitsPadded, 8, 1, false)
       slice = new Uint8Array(sliceBytes).buffer
       break
     }

extensions/essentials/src/binary-to-text.ts

@@ -149,7 +149,7 @@ const execute: OperationExecuteExport = (request) => {
     ))
 
     // Transform unit size (encode)
-    const encodedUnits = transformUnitSize(units, inUnitSize, outUnitSize)
+    const encodedUnits = transformUnitSize(units, inUnitSize, outUnitSize, true)
     const encodedCodePoints = encodedUnits.map(unit => alphabet[unit])
 
     // Apply padding
@@ -195,7 +195,7 @@ const execute: OperationExecuteExport = (request) => {
     encodedUnits = encodedUnits.slice(0, j)
 
     // Transform unit size (decode)
-    const units = transformUnitSize(encodedUnits, outUnitSize, inUnitSize)
+    const units = transformUnitSize(encodedUnits, inUnitSize, outUnitSize, false)
     const buffer = Uint8Array.from(units).buffer
 
     // Add an issue if foreign characters were encountered

extensions/essentials/src/lib/binary.ts

@@ -75,7 +75,8 @@ export const isBigEndianEnvironment = (): boolean =>
 export const transformUnitSize = (
   inputUnits: number[],
   inUnitSize: number,
-  outUnitSize: number
+  outUnitSize: number,
+  forward: boolean
 ): number[] => {
   if (inUnitSize === outUnitSize) {
     return inputUnits
@@ -84,11 +85,18 @@ export const transformUnitSize = (
     return []
   }
 
+  // Some bits may not hold any information when decoding, see issue #40
+  if (!forward) {
+    [inUnitSize, outUnitSize] = [outUnitSize, inUnitSize]
+  }
+
   const commonSize = lcm(inUnitSize, outUnitSize)
   const inUnits = commonSize / inUnitSize
   const outUnits = commonSize / outUnitSize
 
-  const outputUnitsLength = Math.ceil((outUnits * inputUnits.length) / inUnits)
+  const outputUnitsLength = forward
+    ? Math.ceil((outUnits * inputUnits.length) / inUnits)
+    : Math.floor((outUnits * inputUnits.length) / inUnits)
   const outputUnits = new Array<number>(outputUnitsLength)
 
   let remainingInBits, inUnit, moveBits
@@ -104,6 +112,12 @@ export const transformUnitSize = (
       // If the current out unit is full, move to the next one
       if (remainingOutBits === 0) {
         outputUnits[o++] = outUnit
+        // If the remaining input bits contain no information, don't add a null unit (only occurs when decoding)
+        // TODO: This works, but I, anderium, don't like that this `if` executes on each iteration.
+        if (o === outputUnitsLength) {
+          o--
+          break
+        }
         outUnit = 0
         remainingOutBits = outUnitSize
       }