I’m wondering what the best way would be to concatenate multiple CSV files – all of which contain the same header row (they are compatible) – into one CSV file. I know I can do something like (taken from @c42f ’s one-liner ): vcat(CSV.read.(file_names)...) |> CSV.write("one_big_file.csv") But since…

See my answer here on removing the first line from a text file. Don’t parse the data, just copy it blindly to a new file, but skip the first (header) line for everything but the first file.

Very handy @stevengj ! But we still have the issue of merging all these files without loading them into memory, right @yakir12 ? What’s the best way to solve that?

Steve’s nice solution using: Iterators.drop(eachline(input), 1) reads and writes the files line by line. I have just tried it to merge two 47GB files and it seemed to work, taking ~10 min on my PC laptop.

[image] rafael.guerra: Steve’s nice solution using: Iterators.drop(eachline(input), 1) reads and writes the files line by line. After you read the first line to strip off the header, it will be a lot faster to read the file in chunks, as in this example: How to obtain the result of a diff bet…

[image] stevengj: to read the file in chunks In the example linked, the chunks seem to be 32768 bytes long. Why this value?

It’s a power of 2 and a common size for L1 cache.

Thanks Oscar. In my laptop I see this: [L1_L2_L3_cache] Does it mean that I should use a chunk size = 2^18 = 262144 (< 320 KB)?

it probably will be a minor difference, but feel free to do some benchmarks…

@Oscar_Smith , fyi, I’ve observed ~22% speed gains on my laptop (with L1 cache = 320 KB) when using a chunk size of 262_144 bytes instead of 32_768 bytes. In any case, doing it by chunks seemed much faster than doing it line by line. The code used to merge 2 x 37 GB csv files was adapted from Steve’…

This approach worked like a charm!!! so fast! thanks!

Concatenate csv files without loading them

General Usage

stevengj February 12, 2022, 11:33pm 5

After you read the first line to strip off the header, it will be a lot faster to read the file in chunks, as in this example: How to obtain the result of a diff between 2 files in a loop? - #4 by stevengj

Topic		Replies	Views
Skipping a lot of lines in CSV.read() allocates too much memory Performance csv , io	77	2653	February 23, 2024
How can I split large data using a faster and more efficient function (data science)? New to Julia csv	9	890	October 27, 2022
Reading headers of delimited files General Usage csv	1	675	January 3, 2023
CSV.Row very slow for reading files line by line Performance package , csv	0	302	May 9, 2023
Read multiple csv files Performance filesystem	1	1455	July 28, 2020

Concatenate csv files without loading them

Related topics