Java：优化文件定界符以提高子文档的读取速度

Question

Say I have a file that has many subdocuments in it

//file.txt

BEGIN_FILE_1
loremipsumloremipsumloremipsum
loremipsumloremipsum
END_FILE_1

BEGIN_FILE_2
cupcakeipsum
cupcakeipsumcupcakeipsum
END_FILE_2

What kind of delimitation (or some alterate strategy) could be used such that the reads of said subdocuments are fast (ie interpreting the delimitation are fast) but even more importantly, the writing of the subdocument is fast. Note that the container file will be very large (100MB or so).

I am planning to use FileWriter to write the file.

Thanks!

Answer 1

Generally, optimal strategy depends on the context - how many sub-documents is there, will each document be written only once or updated/modified, is size of each subdocument known/at least max size of each subdocument known, which operation prevails (for eac h write operation there would be roughly 10 reads, or the opposite)?

On assumption that subdocuments will be added and read but not modified, optimal strategy may be to use header specifying number of files, and line where each file starts/ends inside your file. Something like - first line always header, then lines 1..N FILE1, N+1..M FILE2, and so on:

NUMBER_OF_FILES FILE1_NAME FILE1_START FILE1_END FILE2_NAME FILE2_START FILE2_END

This would allow read contents of any file by parsing header only and reading directly this file instead of searching for file through the document, and writing would require only modifying the header and writing to the end of file.

If files are modified/overwritten but have fixed size, this strategy may still be useful since overwrite operation would be fast