问题
I'm trying to convert mainframe fixed length file from EBCDIC format to ASCII format. Currently I'm reading file using JZOS api (ZFile) and converting field by field. is it possible to convert without knowing the layout of file (aka COPYBOOK) by just reading entire bytes of a record or line? If so how do handle packed decimals and binary values?
回答1:
is it possible to convert without knowing the layout of file (aka COPYBOOK) by just reading entire bytes of a record or line?
No.
Text fields must be converted from EBCDIC to ASCII. Binary and packed decimal fields must not be converted. If you convert binary fields then you will alter their values, it's possible (likely? certain?) you will destroy their values.
Binary fields coming from a mainframe will be big-endian, you may need to convert these to little endian. +10 in a halfword on a mainframe is x'000A' while on a little endian machine it is x'0A00'.
Packed decimal fields may have implied decimal positions. If your file contains x'12345C' that may represent +123.45 or +12,345. The format of the field tells you how to interpret the data.
You cannot do the conversion without knowing the record layout including field formats.
In my experience, the best way to avoid these difficulties is to preprocess the file on the mainframe, converting all binary and packed decimal fields to text with embedded explicit signs and decimal points. Then the file can safely go through code page (EBCDIC to ASCII in this case) conversion.
Such preprocessing can easily be done with the mainframe SORT utility, which typically excel at data transformations.
来源：https://stackoverflow.com/questions/44682590/ebcdic-to-ascii-conversion-using-java