Issue while flattening the JSON file to CSV in RDP ESG Bulk

Bala Ilango
Bala Ilango LSEG
in Refinitiv Data Platform

Hello there,

I did ESG Bulk extraction. However, when I am trying to flatten the JSON to CSV, few columns still has the json tags. It doesn't seem to flatten the JSON file completely. Appreciate you support to investigate and help on this.

For instance, I tried converting the RFT-ESG-Scores-Full-Init-2021-04-25.jsonl.gz file from JSON to CSV using the following code,

#convert specific json to csv
filedestinationpath = 'C:\\$files\\$ESG\\RDP_BULK\\Results\\'
filename = filedestinationpath + 'RFT-ESG-Scores-Full-Init-2021-04-25' + '.jsonl.gz'
f=gzip.open(filename,'rb')
file_content=f.read()
lines = file_content.splitlines()
df_inter = pd.DataFrame(lines)
df_inter.columns = ['json_element']
df_resolve = df_inter['json_element'].apply(json.loads)
df_final = pd.json_normalize(df_resolve)
resultspth = filedestinationpath + 'RFT-ESG-Scores-Full-Init-2021-04-25' + '.csv'
df_final.to_csv(resultspth, index = False)

It seem to convert, but not all the column. For example, OrganizationName doesn't seem to flatten out completely, it still carries the json tags.

image


Similarly, when I tried "RFT-ESG-Symbology-SEDOL-Init-2021-04-29.jsonl"; few columns as shown in the screengrab below seems to be the issue.

image

Appreciate if you can review and support on this.

Tagged:

Best Answer

  • zoya faberov
    zoya faberov ✭✭✭✭✭
    Answer ✓

    Hello @Bala Ilango,

    Perhaps you may wish to take the same approach further, and double-normalize the fields that contain nested objects on RFT-ESG-Scores:

    For example:

    import gzip
    import pandas as pd
    import json
    #convert specific json to csv
    filedestinationpath = '.\\'
    filename = filedestinationpath + 'RFT-ESG-Scores-Current-init-2021-05-02' + '.jsonl.gz'
    f=gzip.open(filename,'rb')
    file_content=f.read()
    lines = file_content.splitlines()
    df_inter = pd.DataFrame(lines)
    df_inter.columns = ['json_element']
    df_resolve = df_inter['json_element'].apply(json.loads)
    df_resolve
    df_final = pd.json_normalize(df_resolve)
    df_final['ESGOrganization.Names.Name.OrganizationName'] = pd.json_normalize(df_final['ESGOrganization.Names.Name.OrganizationName'].str[0]
    resultspth = filedestinationpath + 'RFT-ESG-Scores-Current-init-2021-05-02' + '.csv'
    df_final.to_csv(resultspth, index = False) 
    df_final

    Resulting in

    image


Answers

  • [Deleted User]
    [Deleted User] Newcomer

    Hi @Bala Ilango,

    May I ask how you got the 'RFT-ESG-Scores-Full-Init-2021-04-25.jsonl.gz' data file in question? was it through Python? If so, would you mind sharing your code (removing identifiable text)?

  • Bala Ilango
    Bala Ilango LSEG

    @jonathan.legrand Yes. I am using Python. Shared the codes via email. Appreciate your support with this. Thanks.

  • zoya faberov
    zoya faberov ✭✭✭✭✭

    Hello @Bala Ilango,

    The columns that do not convert, and retain square brackets and curly braces are nested objects.

    ESGOrganization.Names.Name.OrganizationName is in this case a nested object, implementing array and containing a map, with, potentially, multiple names, for example:

    {"ObjectId":"4295864969;111","StatementDetails":{"OrganizationId":"4295864969","FinancialPeriodEndDate":"2020-12-31T00:00:00.000Z","FinancialPeriodFiscalYear":"2020","FinancialPeriodIsIncomplete":"true"},"ESGOrganization":{"Names":{"Name":{"OrganizationName":[{"OrganizationNormalizedName":"China High Speed Transmission Equipment Group Co Ltd"}]}}},"ESGScores":{"ESGCombinedScore":{"Value":"0.5429413762056495","ValueCalculationDate":"2021-05-01T18:05:32.161Z","ValueScoreGrade":"B-"},"ESGScore":{"Value":"0.5429413762056495","ValueCalculationDate":"2021-05-01T18:05:32.161Z","ValueScoreGrade":"B-"},"EnvironmentPillarScore":{"Value":"0.6736242884250474","ValueCalculationDate":"2021-04-24T17:12:23.608Z","ValueScoreGrade":"B+ "},"ESGResourceUseScore":{"Value":"0.7903225806451613","ValueCalculationDate":"2021-04-03T21:34:07.385Z","ValueScoreGrade":"A-"},"ESGEmissionsScore":{"Value":"0.7258064516129032","ValueCalculationDate":"2021-04-24T17:12:23.608Z","ValueScoreGrade":"B+"},"ESGInnovationScore":{"Value":"0.5","ValueCalculationDate":"2021-04-03T21:34:07.385Z","ValueScoreGrade":"C+"},"SocialPillarScore":{"Value":"0.4414690382081688","ValueCalculationDate":"2021-04-24T17:12:23.608Z","ValueScoreGrade":"C+ "},"ESGWorkforceScore":{"Value":"0.6375","ValueCalculationDate":"2021-04-10T17:37:21.596Z","ValueScoreGrade":"B"},"ESGHumanRightsScore":{"Value":"0.15217391304347827","ValueCalculationDate":"2021-04-03T21:34:07.385Z","ValueScoreGrade":"D"},"ESGCommunityScore":{"Value":"0.4375","ValueCalculationDate":"2021-04-24T17:12:23.608Z","ValueScoreGrade":"C+"},"ESGProductResponsibilityScore":{"Value":"0.3484848484848485","ValueCalculationDate":"2021-04-10T17:37:21.596Z","ValueScoreGrade":"C"},"GovernancePillarScore":{"Value":"0.4760119460883173","ValueCalculationDate":"2021-05-01T18:05:32.161Z","ValueScoreGrade":"C+ "},"ESGManagementScore":{"Value":"0.5246305418719212","ValueCalculationDate":"2021-05-01T18:05:32.161Z","ValueScoreGrade":"B-"},"ESGShareholdersScore":{"Value":"0.46798029556650245","ValueCalculationDate":"2021-04-10T17:37:21.596Z","ValueScoreGrade":"C+"},"ESGCsrStrategyScore":{"Value":"0.24496644295302014","ValueCalculationDate":"2021-04-10T17:37:21.596Z","ValueScoreGrade":"D+"},"ESGCControversiesScore":{"Value":"1.0","ValueCalculationDate":"2021-04-03T21:34:07.385Z","ValueScoreGrade":"A+"}},"DiversityAndInclusionScores":{"ControversiesScore":{"Value":null,"ValueCalculationDate":null},"DiversityScore":{"Value":null,"ValueCalculationDate":null},"InclusionScore":{"Value":null,"ValueCalculationDate":null},"PeopleDevelopmentScore":{"Value":null,"ValueCalculationDate":null},"Score":{"Value":null,"ValueCalculationDate":null}}}

    Therefore, it will not fully flatten into CSV, without loosing meaning. If you are ok with only the first name represented, you can do something like

    df = pd.json_normalize(df_resolve)
    df['ESGOrganization.Names.Name.OrganizationName'] = df['ESGOrganization.Names.Name.OrganizationName'].str[0]

    And similarly, you can select the key and the value and insert them into dataframe as separate columns, removing the hierarchical column, but I think you would like to preserve the structure in this case, as it enables you to retain the meaning.



  • Bala Ilango
    Bala Ilango LSEG
    https://community.developers.refinitiv.com/discussion/comment/77762#Comment_77762

    @zoya faberov Can you please help me with my other part of the question on best practice to flatten the ESG bulk Symbology files.

  • Bala Ilango
    Bala Ilango LSEG
    https://community.developers.refinitiv.com/discussion/comment/78190#Comment_78190

    @zoya faberov When I am running the following line in the code.

    df_final['ESGOrganization.Names.Name.OrganizationName'] = pd.json.normalize(df_final['ESGOrganization.Names.Name.OrganizationName'].str[0])

    I am getting the following error: 

    AttributeError: module 'pandas' has no attribute 'json'
  • zoya faberov
    zoya faberov ✭✭✭✭✭
    https://community.developers.refinitiv.com/discussion/comment/78191#Comment_78191

    @Bala Ilango

    try pd.json_normalize

    pd.json.normalize appears to be a typo

  • zoya faberov
    zoya faberov ✭✭✭✭✭

    Hello @Bala Ilango,

    To save ESG Bulk symbology files in CSV format, you may also find example ESGBulkToCSV on GutHub useful.

  • anita.varma
    anita.varma LSEG

    Update on this post..

    Flat CSV files are now available for ESG Bulk files.

Categories