Rearrange bytes from split format to normal format for Unsigned Column types #357
Conversation
modules/rntuple.mjs
Outdated
| const byteSize = getTypeByteSize(coltype), | ||
| splitView = new DataView(blob.buffer, blob.byteOffset, blob.byteLength), | ||
| count = blob.byteLength / byteSize, | ||
| fullBuffer = new ArrayBuffer(blob.byteLength), |
There was a problem hiding this comment.
Rather than fullBuffer/fullBytes I'd call them outBuffer/outBytes
modules/rntuple.mjs
Outdated
| for (let b = 0; b < byteSize; ++b) { | ||
| const splitIndex = b * count + i, | ||
| byte = splitView.getUint8(splitIndex), | ||
| writeIndex = i * byteSize + (LITTLE_ENDIAN ? b : byteSize - 1 - b); |
There was a problem hiding this comment.
You can assume LITTLE_ENDIAN is true (if that were not the case, several other things would break at the moment and there is no big endian platform I'm aware of where you would use a browser anyway)
modules/rntuple.mjs
Outdated
| coltype === ENTupleColumnType.kSplitReal64 ? ENTupleColumnType.kReal64 : | ||
| coltype === ENTupleColumnType.kSplitIndex64 ? ENTupleColumnType.kIndex64 : | ||
| coltype === ENTupleColumnType.kSplitUInt32 ? ENTupleColumnType.kUInt32 : | ||
| coltype; |
There was a problem hiding this comment.
You should assert that coltype is one of those listed above; maybe use a switch instead of a chain of ternary operators and have a throw in the default case
There was a problem hiding this comment.
Please check @silverweed the commit 2 I did all your suggested changes
modules/rntuple.mjs
Outdated
| numValues = byteSize ? blob.byteLength / byteSize : undefined; | ||
| reader = new RBufferReader(processedBlob), | ||
| values = [], | ||
| numValues = byteSize ? processedBlob.byteLength / byteSize : undefined; |
There was a problem hiding this comment.
I don't think it makes sense to continue running if byteSize is not defined - you should throw rather than running the loop on processedBlob.byteLength
…g logic Suggested Changes done
In this commit I can successfuly deserialise
unsigned slip columnsIn the below image you can see the test I ran for compressed
ntpl001_staff.rootfile where I checked for flag what was unsigned integer split column which is successfully deserialised but for other fields (e.g Nation) which is not unsigned I get the the error as expected.