20,073,124 array - How would you do it ?

[argumentum]

I have a CSV file.
Is 3.12 GB (3,357,068,152 bytes). 
FileReadToArray() loaded ( just the count ) and UBound() say that is 20073124 and that just happens to be a tiny bit more that the max of 16 million elements an array can hold.

How can a file that size be read, chop in @CRLF chunks, filtered ( If Not StringInStr($aFile[$n], ".au3") Then ContinueLoop ) and write to another file the resultant ?

Anything goes: SQLite, memory, etc. .

Thanks

[Nine]

Read it line by line.  First thing that comes to mind...

[argumentum]

35 minutes ago, Nine said:

First thing that comes to mind...

First thing that came to my mind was to read up to X position, split and work on that, then read the next until done 🤷‍♂️

Line by line would be kind of slow I think

Edited yesterday at 04:28 AM by argumentum

[jchd]

Import in SQLite and apply any SQL magic massage needed to your data. You can just use the SQLite CLI (the command-line interface) with a memory (default) or disc-based database. You can define row and column separators if needed, .import the file as --csv, apply SQL to filter out unwanted rows, rewrite the resultant table to csv or whatever format you wish. If this operation turns to routine, store all these commands in a file and run it with the CLI.

Use a recent sqlite3.exe, it's now full of rich features.

Else slap a good regex in the face of the file!

Edited 19 hours ago by jchd

[jchd]

11 hours ago, argumentum said:

FileReadToArray() loaded ( just the count ) and UBound() say that is 20073124 and that just happens to be a tiny bit more that the max of 16 million elements an array can hold.

IIRC the 16M limit is just an order of magnitude; the actual limit significantly depends on the volume of data.

[Nine]

8 hours ago, argumentum said:

Line by line would be kind of slow I think

In fact it would be faster.  You do not need to create a useless intermediate array.  Reading a bunch of lines won't get it faster either.  The file is preemptively in memory buffers by the OS.  I remember trying to use an overlapped read to perform tasks while reading the file and it ends up with a very little gain but made the script way more complex.  Try it and you will be gladly surprised.

Edited 16 hours ago by Nine

[argumentum]

9 hours ago, Nine said:

Try it and you will be gladly surprised.

True. I did and yes, it was good