# Can't read a CSV file with empty fields into a DataFrame
**URL:** https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276
**Category:** Data
**Tags:** dataframes, csv
**Created:** [April 29, 2022, 10:32pm UTC](https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276 "2022-04-29T22:32:56Z")
**Posts on this page:** 8
**Page:** 1
### Author:  [@Gregstrq](https://discourse.julialang.org/u/Gregstrq)
#### Post date: [April 29, 2022, 10:32pm UTC](https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276/1 "2022-04-29T22:32:56Z")
I am trying to read a CSV file into a DataFrame. It clearly has some empty fields, and I get an error
> ERROR: MethodError: Cannot `convert` an object of type Missing to an object of type String
I have tried to redefine the conversion for missing values:
```julia
Base.convert(::Type{String}, ::Missing) = "missing"
```
But now I get a different error
> ERROR: ArgumentError: missing is not a valid key for type String
**What is the standard way to deal with the empty string fields of a CSV file?**
---
### Author:  [@pdeffebach](https://discourse.julialang.org/u/pdeffebach)
#### Post date: [April 29, 2022, 10:45pm UTC](https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276/2 "2022-04-29T22:45:42Z")
I can’t replicate
```julia
julia> using DataFrames, CSV, Chain;
julia> t = """
a, b
"a1", "b1"
, "b2"
""";
julia> @chain t begin
IOBuffer
CSV.read(DataFrame)
end
2×2 DataFrame
Row │ a b
│ String3? String3
─────┼───────────────────
1 │ a1 b1
2 │ missing b2
```
Can you show us a bit of the file you are using?
---
### Author:  [@Gregstrq](https://discourse.julialang.org/u/Gregstrq)
#### Post date: [April 30, 2022, 9:29am UTC](https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276/3 "2022-04-30T09:29:03Z")
You can download it here: [https://www-nds.iaea.org/nuclearmoments/magn\_mom\_recomm.csv](https://www-nds.iaea.org/nuclearmoments/magn_mom_recomm.csv)
It is a file with the data on nuclear isotopes.
---
### Author:  [@nilshg](https://discourse.julialang.org/u/nilshg)
#### Post date: [April 30, 2022, 10:00am UTC](https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276/4 "2022-04-30T10:00:59Z")
There’s something funky going on on row 121 of the file:
```julia
julia> using CSV, DataFrames, Downloads
julia> CSV.read(Downloads.download("https://www-nds.iaea.org/nuclearmoments/magn_mom_recomm.csv"), DataFrame; limit = 120)
120×12 DataFrame
Row │ z n.n+n.z symbol energy [keV] halflife spin magnetic dipole [nm] method description nsr journal indc
│ Int64 Int64 String3 Int64 String15 String7 String31 String15 String String31 String String15
─────┼──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
1 │ 0 1 NN 0 10.2 m 1/2+ -1.9130427(5) R Re-evaluated by N.J. Stone (2013… 2014Ol01 Chin. Phys. C38 090001 (2014) indc-nds-0794
2 │ 1 1 H 0 stable 1/2+ +2.792847351(9) R Re-evaluated by N.J. Stone (2013… 2014Ol01 Chin. Phys. C38 090001 (2014) indc-nds-0794
```
Opening the plain text file it looks like the next row has a value of `"PRL 96` in the `journal` column, so maybe that’s the culprit that throws parsing off?
---
### Author:  [@Gregstrq](https://discourse.julialang.org/u/Gregstrq)
#### Post date: [April 30, 2022, 10:10am UTC](https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276/5 "2022-04-30T10:10:16Z")
Ok. Yesterday in the evening I was trying to load the file as `CSV.File("magn_mom_recomm.csv")|>DataFrame` and it gave me the cryptic errors I described.
Now, I have tried to do `load("magn_mom_recomm.csv")|>DataFrame`, and I see the error is about the row 121 indeed. It looks like an extra double quotes, which throws parsing off the course.
---
### Author:  [@pdeffebach](https://discourse.julialang.org/u/pdeffebach)
#### Post date: [April 30, 2022, 10:35am UTC](https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276/6 "2022-04-30T10:35:43Z")
Please file an issue with CSV.jl. It should probably be smart enough to throw this as an error
---
### Author:  [@rocco\_sprmnt21](https://discourse.julialang.org/u/rocco_sprmnt21)
#### Post date: [May 9, 2022, 8:44pm UTC](https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276/7 "2022-05-09T20:44:01Z")
try using kwarg `quoted = false`
PS
if you use the kwarg quotechar set to something other than ’ " ’ the df is built regularly (it seems)
```julia
julia> df=CSV.read("magn_mom_recomm2.csv",quotechar='^',DataFrame)
2155×12 DataFrame
Row │ z n.n+n.z symbol "energy [keV]" halflife spin "magnet ⋯
│ Int64 Int64 String3 String15? String15? String15? String3 ⋯
──────┼──────────────────────────────────────────────────────────────────────────────
1 │ 0 1 NN 0 "10.2 m" 1/2+ -1.9130 ⋯
2 │ 1 1 H 0 stable 1/2+ +2.7928
3 │ 1 2 H 0 stable 1+ +0.8574
4 │ 1 3 H 0 "12.33 y" 1/2+ +2.9789
5 │ 2 3 He 0 stable 1/2+ -2.1276 ⋯
6 │ 3 6 Li 0 stable 1+ +0.8220
7 │ 3 7 Li 0 stable 3/2- +3.2564
8 │ 3 8 Li 0 "840 ms" 2+ +1.6535
9 │ 3 9 Li 0 "178 ms" 3/2- 3.43666 ⋯
10 │ 3 11 Li 0 "8.75 ms" 3/2- +3.6711
11 │ 4 9 Be 0 stable 3/2- -1.1774
12 │ 4 11 Be 0 "13.8 s" 1/2+ -1.6816
13 │ 5 8 B 0 "0.77 s" 2+ 1.0355( ⋯
14 │ 5 10 B 718 "0.707 ns" 1+ +0.63(1
15 │ 5 10 B 0 stable 3+ 1.80046
16 │ 5 11 B 0 stable 3/2- 2.68837
17 │ 5 12 B 0 "20.2 ms" 1+ +1.003( ⋯
18 │ 5 13 B 0 "17.3 ms" 3/2- +3.1778
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋱
2138 │ 93 237 Np 60 "67 ns" 5/2- +1.85(2
2139 │ 93 239 Np 75 "1.39 ns" 5/2- +2.2(4)
....
2154 │ 99 254 Es 84 "39.3 h" 2+ 2.90(7)
2155 │ 102 253 No 0 "1.62 m" (9/2-) -0.53(8
6 columns and 2119 rows omitted
```
---
### Author:  [@ymh](https://discourse.julialang.org/u/ymh)
#### Post date: [May 16, 2022, 12:06am UTC](https://discourse.julialang.org/t/cant-read-a-csv-file-with-empty-fields-into-a-dataframe/80276/8 "2022-05-16T00:06:08Z")
omg I’ve been having issues with this (i’m working with messy text data), and setting quoted=false did the trick