### Author:  [@tpoterba](https://discuss.hail.is/u/tpoterba)
#### Post date: [August 15, 2018, 9:28pm UTC](https://discuss.hail.is/t/entry-fields-empty-after-import-vcf-from-1000-genomes-data/608/2 "2018-08-15T21:28:31Z")
> [@greg-botwin](#):
>
> gs://genomics-public-data/1000-genomes-phase-3/vcf-20150220/ALL.chr1.phase3\_shapeit2\_mvncall\_integrated\_v5a.20130502.genotypes.vcf
this VCF is 4.1; we are usually seeing v4.2 vcfs. Maybe we should throw an error on 4.1.
The primary probem here is that the FORMAT fields aren’t declared in the header!
---
### Author:  [@greg-botwin](https://discuss.hail.is/u/greg-botwin)
#### Post date: [August 15, 2018, 10:19pm UTC](https://discuss.hail.is/t/entry-fields-empty-after-import-vcf-from-1000-genomes-data/608/3 "2018-08-15T22:19:56Z")
Thanks Tim for your superfast reply!!
I just checked out the header for the VCF file and I think it does have some FORMAT information in the header. What is the minimum that import\_vcf needs?
> bcftools view -h ALL.chr1.phase3\_shapeit2\_mvncall\_integrated\_v5a.20130502.genotypes.vcf.gz
> ##fileformat=VCFv4.1
> …
> ##FORMAT=\
I naively tried to load the file using `import_matrix_table` and manually specifying the rows, but it did not work.
```
mt = hl.import_matrix_table('gs://genomics-public-data/1000-genomes-phase-3/vcf-20150220/ALL.chr1.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf',
row_fields={'locus': hl.tlocus,
'alleles': hl.tarray,
'rsid': hl.tstr,
'qual': hl.tfloat64,
'filters': hl.tset,
'info': {
'CIEND': hl.tarray(hl.tint32),
'CIPOS': hl.tarray(hl.tint32),
'CS': hl.tstr,
'END': hl.tint32,
'IMPRECISE': hl.tbool,
'MC': hl.tarray(hl.tstr),
'MEINFO': hl.tarray(hl.tstr),
'MEND': hl.tint32,
'MLEN': hl.tint32,
'MSTART': hl.tint32,
'SVLEN': hl.tarray(hl.tint32),
'SVTYPE': hl.tstr,
'TSD': hl.tstr,
'AC': hl.tarray(hl.tint32),
'AF': hl.tarray(hl.tfloat),
'NS': hl.tint32,
'AN': hl.tint32,
'EAS_AF': hl.tarray(hl.tfloat),
'EUR_AF': hl.tarray(hl.tfloat),
'AFR_AF': hl.tarray(hl.tfloat),
'AMR_AF': hl.tarray(hl.tfloat),
'SAS_AF': hl.tarray(hl.tfloat),
'DP': hl.tint32,
'AA': hl.tstr,
'VT': hl.tarray(hl.tstr),
'EX_TARGET': hl.tbool,
'MULTI_ALLELIC': hl.tbool}
},
row_key = ['locus', 'alleles']
)
```
> TypeError: import\_matrix\_table: parameter ‘row\_fields’: expected dict[str, (hail.expr.types.HailType or (str))], found dict: {‘locus’:…
Thanks!
---
### Author:  [@tpoterba](https://discuss.hail.is/u/tpoterba)
#### Post date: [August 16, 2018, 3:21am UTC](https://discuss.hail.is/t/entry-fields-empty-after-import-vcf-from-1000-genomes-data/608/4 "2018-08-16T03:21:12Z")
It may be possible to import a VCF using import\_matrix\_table with some more wrangling (you’ll need to set `entry_type` to a string or call, and the `info` field to a string (the VCF info field is a bit special and crazy, the `import_vcf` has a bunch of specialized code).
However, that’s definitely not something I’d recommend! I think this may just be some bug inside Hail or the way we call out to HTSJDK, so I’ve made an issue to track that - [https://github.com/hail-is/hail/issues/4160](https://github.com/hail-is/hail/issues/4160)
I’m on vacation the rest of this week but randomly assigned Jon to take a look.
---
### Author:  [@greg-botwin](https://discuss.hail.is/u/greg-botwin)
#### Post date: [August 16, 2018, 4:39pm UTC](https://discuss.hail.is/t/entry-fields-empty-after-import-vcf-from-1000-genomes-data/608/5 "2018-08-16T16:39:50Z")
Thanks Tim. Enjoy your vacation! Very thankful for the support you and your team provide!
---
### Author:  [@jbloom](https://discuss.hail.is/u/jbloom)
#### Post date: [August 17, 2018, 7:23pm UTC](https://discuss.hail.is/t/entry-fields-empty-after-import-vcf-from-1000-genomes-data/608/6 "2018-08-17T19:23:46Z")
Our colleague Liam has made 1000 genomes phase 3 matrix tables publicly available at:
```auto
https://console.cloud.google.com/storage/browser/hail-datasets/hail-data/
```
Not sure what’s going on yet, but that should allow you to experiment in the meantime!
---
### Author:  [@jbloom](https://discuss.hail.is/u/jbloom)
#### Post date: [August 17, 2018, 7:55pm UTC](https://discuss.hail.is/t/entry-fields-empty-after-import-vcf-from-1000-genomes-data/608/7 "2018-08-17T19:55:26Z")
So Liam found that the genomics-public-data VCF hosted by Google is missing the line `##FORMAT=` in the header. He created the matrix tables above using the files directly hosted by EBI:
ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/release/20130502/
---
### Author:  [@tpoterba](https://discuss.hail.is/u/tpoterba)
#### Post date: [August 18, 2018, 1:52pm UTC](https://discuss.hail.is/t/entry-fields-empty-after-import-vcf-from-1000-genomes-data/608/8 "2018-08-18T13:52:44Z")
can we give Google a fixed VCF?
---
### Author:  [@greg-botwin](https://discuss.hail.is/u/greg-botwin)
#### Post date: [August 20, 2018, 4:15pm UTC](https://discuss.hail.is/t/entry-fields-empty-after-import-vcf-from-1000-genomes-data/608/9 "2018-08-20T16:15:06Z")
@jbloom @tpoterba thanks again for looking into this! The amazing support your team provides for hail is one of the many reasons I am excited to learn more about hail and share it with my colleagues (currently transitioning from array data using plink to ngs).
---
### Author:  [@tpoterba](https://discuss.hail.is/u/tpoterba)
#### Post date: [August 21, 2018, 12:15am UTC](https://discuss.hail.is/t/entry-fields-empty-after-import-vcf-from-1000-genomes-data/608/10 "2018-08-21T00:15:37Z")
I’ve sent an email to GCP support to see if we can get this fixed for posterity.