# Writing larger than expected binary files

**URL:** <https://discourse.julialang.org/t/writing-larger-than-expected-binary-files/62600>\
**Category:** General Usage\
**Tags:** fortran, binaryio\
**Created:** [June 9, 2021, 2:35am UTC](https://discourse.julialang.org/t/writing-larger-than-expected-binary-files/62600 "2021-06-09T02:35:02Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![k945](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/k945/32/8906_2.png) [@k945](https://discourse.julialang.org/u/k945)\
**Post date:** [June 9, 2021, 2:35am UTC](https://discourse.julialang.org/t/writing-larger-than-expected-binary-files/62600/1 "2021-06-09T02:35:02Z")

</div>

When using FortranFiles, I was wondering why writing a new file, with the same data and structure as the original, produces a file whose size is significantly larger (~925MB) than the original (~25MB). Is there a compression option that I should be using when writing binary files?

MWE

```julia
using FortranFiles, DimensionalData

function getvar(fn, vname)
    f = FortranFile(fn, convert = "big-endian")

    # We need to define these variables due to scoping within the while loop
    slab, ifv, hdate, xfcst, map_source, field, units, desc,
    xlvl, nx, ny, iproj, projrecs, is_wind_earth_rel, meta = [nothing for _ in 1:15]

    levels = Float32[]

    while !eof(f)
        ifv = read(f, Int32)
        hdate, xfcst, map_source, field, units, desc, xlvl, nx, ny, iproj =
            read(f, FString{24}, Float32, FString{32}, FString{9},
                FString{25}, FString{46}, Float32, Int32, Int32, Int32)

        projrecs = projectionrec(Val(Int(iproj)), f)
        is_wind_earth_rel = read(f, Int32)

        if String(trim(field)) == vname
            push!(levels, xlvl)
            if slab == nothing
                slab = read(f, (Float32, nx,ny))
            else
                slab = cat(slab, read(f, (Float32, nx,ny)), dims = 3)
            end
            meta = (ifv = ifv, hdate = hdate, xfcst = xfcst, map_source = map_source,
                units = units, desc = desc, iproj = iproj, projrecs = projrecs,
                is_wind_earth_rel = is_wind_earth_rel)
        else
            read(f)
        end
    end
    close(f)

    DimArray(slab, DimensionalData.formatdims(slab, (X(1:nx), Y(1:ny), Z(levels))), (), Symbol(vname), meta)
end

function Base.write(fn::String, T::Tuple)
    isfile(fn) ? rm(fn) : nothing
    f = FortranFile(fn, "w", convert = "big-endian")
    nx = convert(Int32, size(T[1], X))
    ny = convert(Int32, size(T[1], Y))
    xlvl = convert(Array{Float32}, val(T[1], Z))

    for eachlev in axes(T[1], 3)
        for (idx, eachvar) in enumerate(T)
            meta = metadata(eachvar)
            field = FString(9, String(name(eachvar)))
            projrecs = values(meta[:projrecs])

            write(f, meta[:ifv])
            write(f, meta[:hdate], meta[:xfcst], meta[:map_source], field, meta[:units], meta[:desc], xlvl[eachlev], nx, ny, meta[:iproj])
            write(f, projrecs...)
            write(f, meta[:is_wind_earth_rel])
            write(f, data(eachvar))
        end
    end
    close(f)
end

function projectionrec(::Val{0}, f)
    startloc, startlat, startlon, deltalat, deltalon, earth_radius = read(f, FString{8}, Float32, Float32, Float32, Float32, Float32)
    (startloc = startloc, startlat = startlat, startlon = startlon, deltalat = deltalat, deltalon = deltalon, earth_radius = earth_radius)
end

# FILE can be downloaded from https://cloudstor.aarnet.edu.au/plus/s/rolTL7Phye758q1
fn = "FILE"
pres = getvar(fn, "PRES");
TT = getvar(fn, "TT");
VV = getvar(fn, "VV");
UU = getvar(fn, "UU");
GHT = getvar(fn, "GHT");
SPECHUMD = getvar(fn, "SPECHUMD");

write("testing", (TT, VV, UU, GHT, SPECHUMD, pres))

```

---

<div class="post-metadata">

**Author:** ![k945](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/k945/32/8906_2.png) [@k945](https://discourse.julialang.org/u/k945)\
**Post date:** [June 12, 2021, 3:23am UTC](https://discourse.julialang.org/t/writing-larger-than-expected-binary-files/62600/2 "2021-06-12T03:23:59Z")

</div>

The large file size was due to writing the entire array, through each write cycle, rather than the relevant part. Updating the inner loop solved the problem

```julia
for (idx, eachvar) in enumerate(T)
    A = eachvar[:,:,eachlev]

    meta = metadata(A)
    field = FString(9, String(name(eachvar)))
    projrecs = values(meta[:projrecs])

    write(f, meta[:ifv])
    write(f, meta[:hdate], meta[:xfcst], meta[:map_source], field, meta[:units], meta[:desc], xlvl[eachlev], nx, ny, meta[:iproj])
    write(f, projrecs...)
    write(f, meta[:is_wind_earth_rel])
    write(f, data(A))
end

```