Monday, March 16, 2020

Python to Tableau Hyper (P2H)

Sometimes, we could come across a requirement to get data from a Cloud based API (Application Programming Interface; I know the abbreviation!) providing JSON data (JavaScript Object Notation .. sigh!!) to be analyzed directly in a visualization tool like Tableau.

You may or may not want to store the data in an intermediate data store (like Database or flat files) but would prefer to analyze this directly in the Tableau BI tool. Also, you may need the data to be extracted and refreshed in its entirety or incrementally from time to time and have it published to Tableau Server (if you have one!) so your users could make use of the same.

How or what to do for such a requirement?

Well if you are aware of programming languages like Python, then you are aware that this could be done easily.

You would have to write some code for it though to get the data from Cloud API.


But where to store the data? 

Tableau is friendly with and uses a proprietary Hyper File Format (.hyper files) which it could connect to directly to analyze the underlying data. This could also be published directly to Tableau Server. Hyper files are essentially mini data instances of PostgreSQL database which could hold data and also provide SQL features like Insert, Update, Delete, etc.


Sounds friendly enough? This is how we do it in a simple flow.


Okay this may look like a cheesy graphic, but does explains what we are doing here.

Moreover the aim here is to give you sample code and installation steps, so you could modify the same to connect to any Cloud API and get your required data in Tableau Hyper for your analysis.

So how about doing this?

There are two main things which are involved here.
  1. Setting up the Environment and Installing the code (use the sample code provided here)
  2. Understanding the code, so you could modify this to your requirements.

& here we begin...

-->For setting up the environment and installing the code, four main pieces are required.
  1. Install Python and dependencies like requests, faker (optional) 
  2. Install tableauhyperapi dependency for Python link here along with Microsoft Visual C++ Runtime
  3. Install Visual Studio Code with Pylint extension
  4. Install Tableau Desktop
I am using Faker to generate random data. Its use is optional. 

I have done my setup on Windows and followed standard installation steps for each of the components above. I believe my code could be easily ported to Linux environment too, but I haven't really tested there, so its your game there. Also, I prefer installation of tableauhyperapi dependency using .whl file for Python.

A typical installation steps are given in Readme document along with the attached source code.

You are responsible for licenses of above products as it may be required.


--> Understanding the code, so you could modify this to your requirements is a major step. Let's try digging in the code by first running it and then understanding it file by file.

To run the code open a command prompt to the extracted code directory and issue a command in code_extracted_path directory like path_to_python_exe\python.exe Main.py -h. For e.g.

D:\P2H_v1>c:\Python\Python38\python.exe Main.py -h

You would get output like below


Let's try to understand this output, the arguments which could be passed to the P2H Python Script.
  1. help: -h or --help to see this help message
  2. output file: --output to specify the output file .hyper or .tdsx. You know about .hyper already, .tdsx is one step above .hyper file as it could contain Data Model as well joining different tables inside the hyper file. Since, we could not generate Data Model using this Python Script, the .tdsx file should be pre-existing so that the data inside it could be updated and existing data model could be used.
  3. API URL: --sampleapiurl is the URL for the API endpoint, could be specified here or a defaul could be set inside the Main.py script.
  4. API key: --sampleapikey is the authentication key which would be base64 encoded and sent to API in the request header for Basic Authorization.
  5. Load Type: --load is used for specifying type of load FULL or INC
  6. Incremental Date: --incdate is used for specifying the date from which data would be pulled from API. This needs configuration inside the script and API should support such method.
  7. Incremental Period: --incperiod is used if you want to specify period like Last Week, Last Month, Last Year, etc. instead of incdate for pulling the data. This helps in automating batch pulls.
  8. Email To: --emailto specifies the email address of the recipients who would receive email after the script has completed processing. This requires mail server configuration inside the script.
  9. Package XML: --packagexml option is different. Remember we could have .tdsx file as output? So if you have the data model xml .tds file from inside .tdsx file from somewhere, you could specify path to that xml file with this option and the script would package it along with .hyper file to create .tdsx for you.
  10. Backup: --backup, yes or no if you want to backup existing output file before doing the updates.
  11. Backup Path: --backuppath specifies the path for backup file, to be used with --backup option.
  12. Log: --log is used to specify the Python script log file location. If nothing is specified then log output is generated on screen.

The minimum input parameters that are needed are --output and --sampleapikey for the script to run and process the data from Sample API used in the script. For e.g. use below to generate sampleapi.hyper file.

D:\P2H_v1>c:\Python\Python38\python.exe Main.py --output sampleapi.hyper --sampleapikey any_text_for_this



So in the code_extracted_path directory you would now see the sampleapi.hyper file which you could open with Tableau Desktop to analyze the data inside.


Now the code has run and there are a lot more things to observe, let's dig further.

Dig into the code, yummm me likey!

First things you see with running the sample code is obviously the log which is generated on-screen and the sampleapi.hyper file which is produced. Inside the sampleapi.hyper file you see two tables Users and Resource which hold the real data from the API. You also see a Extract_Log table which records statistics on various data pulls for each run for each table. This could prove useful in analyzing what happened when the things go south, sigh!

So what are the components making this data pull possible?

Obviously these are the two Python scripts Main.py and SampleDataSchema.py working with tableauhyperapi which we installed before and a Sample API (I used https://reqres.in/api) from where we are getting the data.

--> Let's look at Main.py and try to understand what's happening here

Warning! Long code ahead!!

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
#Basic Imports
import sys
import os
import argparse
import datetime
import time
import shutil
import zipfile

#For Email
import smtplib
import email

#For working with REST APIs
import requests
import base64

#For writing Tableau Hyper File
import tableauhyperapi

#Custome import for Sample API Data Schema
import SampleDataSchema

#Global Variables
##Below variables are Input as Arugment from Command Line. Values defined here would be overridden by Command Line
_gv_OutputHyperFile = None                      
_gv_OutputPackageFile = None
_gv_PackageXMLFile = None
_gv_SampleAPIURL = None
_gv_SampleAPIKey = None
_gv_SampleAPIIncDate = None
_gv_LogFile = None
_gv_Backup = None
_gv_BackupFilePath = None
_gv_EmailMessageTo = None

##Modify below variables to change the script behaviour or output
_gv_ApplicationBanner = "Python to Tableau Hyper (P2H)" #Application Banner (appears in Log and Email)
_gv_SampleAPIContentType = "application/json"           #Content Type passed in header for Request to REST API
_gv_SampleAPITimeOut = None                             #Time out in seconds for REST API
_gv_SampleAPIMaxRequestsPerSec = 10                     #Maximum no. of requests made per second made to API
_gv_TabHyperLoadSchema = "Extract"                      #Name of main output schema in Hyper File
_gv_TabHyperStageTablesSuffix = "_Stage"                #Name of Stage tables in Hyper Files.
_gv_TabHyperStageUsePersistentTables = False            #Should Stage Table should be dropped after each load
_gv_TabHyperExtractLogTable = "Extract_Log"             #Name of Extract Log Table inside Hyper File
_gv_EmailServer = None                                  #Email Server full name or ip address
_gv_EmailMessageFrom = None                             #Email Address From
_gv_ArchiveUnarchiveTempPath = "_temp"                  #Temporary Extract path for .tdsx package file
_gv_TabHyperSendUsageStats = tableauhyperapi.Telemetry.DO_NOT_SEND_USAGE_DATA_TO_TABLEAU #Or tableauhyperapi.Telemetry.SEND_USAGE_DATA_TO_TABLEAU for sending Usage Statistics to Tableau

#Variables used in calculations, do not modify
_gv_SampleAPIAuthorization = None
_gv_SampleAPIHeader = None
_gv_SampleAPIBaseURIS = None
_gv_SampleAPIResponse = None
_gv_SampleAPICookies = None
_gv_SampleAPIReqCounter = {"LastAccessed":"", "RequestCount":0}
_gv_LogFilePath = None
_gv_EmailMessage = None
_gv_TabHyperProcess = None
_gv_TabHyperCon = None

#Function to Parse Input Arguments
def ParseArguments():
    global _gv_SampleAPIURL, _gv_SampleAPIKey, _gv_LogFilePath, _gv_SampleAPIIncDate, _gv_OutputHyperFile, _gv_OutputPackageFile, _gv_PackageXMLFile, _gv_Backup, _gv_BackupFilePath, _gv_EmailMessage, _gv_EmailMessageTo, _gv_ArchiveUnarchiveTempPath
    
    #Start of Email Message writing
    _gv_EmailMessage = "----------------------------------------------------\n"
    _gv_EmailMessage += _gv_ApplicationBanner + "\n"
    _gv_EmailMessage += "----------------------------------------------------\n\n"

    _gv_EmailMessage = "Arguments passed: \n"

    ##Parse Arguments - Modify here Input Arguments
    parser = argparse.ArgumentParser()
    parser.add_argument("--output", required=True, help="Full Path to Output File .hyper or .tdsx (.tdsx file should exist); e.g. c:\\new_folder\\SampleAPI_Output_file.hyper", type=str)
    parser.add_argument("--sampleapiurl", required=False, default="https://reqres.in/api", help="SampleAPI API URL, optional otherwise uses Default", type=str)
    parser.add_argument("--sampleapikey", required=True, help="Sample API Key", type=str)
    parser.add_argument("--load", required=False, default="Full", help="Load Type Full or Inc (Incremental); Default: Full", type=str)
    parser.add_argument("--incdate", required=False, help="[Required with load = Inc] Date (YYYY-MM-DD) to extract data from", type=str)
    parser.add_argument("--incperiod", required=False, help="[Required with load = Inc] Period to extract data from e.g. WTD, MTD, YTD, LastWeek, LastMonth, LastYear", type=str)
    parser.add_argument("--emailto", required=False, help="[Optional] (Requires E-Mail Config inside script) Comma separated email addresses", type=str)
    parser.add_argument("--packagexml", required=False, help="[Optional] Full path to .tds XML file to package with .hyper output file; e.g. c:\\new_folder\\SampleAPI_package_XML_file.tds", type=str)
    parser.add_argument("--backup", required=False, help="[Optional] Backup Pre-Existing Output File - Yes or No", type=str)
    parser.add_argument("--backuppath", required=False, help="[Optional] Full Directory Path for Backup, if not provided then Output directory is used", type=str)
    parser.add_argument("--log", required=False, help="[Optional] Full Path to Log File", type=str)
    args = parser.parse_args()

    #Check for Email
    if args.emailto != None:
        if _gv_EmailServer == None or _gv_EmailMessageFrom == None:
            parser.error("The --emailto parameter requires E-mail config (Server, From) to be done inside the script.")
        else:
            _gv_EmailMessageTo = args.emailto
    
    _gv_SampleAPIURL = args.sampleapiurl
    _gv_EmailMessage += "\n Sample API URL: " + str(_gv_SampleAPIURL)
    _gv_SampleAPIKey = args.sampleapikey
    _gv_EmailMessage += "\n Sample API Access Key: " + str(_gv_SampleAPIKey)
    _gv_LogFilePath = args.log
    _gv_EmailMessage += "\n Log File Path: " + str(_gv_LogFilePath)
    
    #Check for Incremental Date input parameter
    if str(args.load).lower() == "inc":
        if args.incdate == None and args.incperiod == None:
            parser.error("The --load=Inc requires --incdate or --incperiod parameter to have a value passed")
        else:
            if args.incdate != None:
                try:
                    _gv_SampleAPIIncDate = datetime.datetime.strptime(args.incdate, "%Y-%m-%d")
                    _gv_EmailMessage += "\n Load Type: INC"
                    _gv_EmailMessage += "\n Inc Date: " + str(_gv_SampleAPIIncDate)
                except Exception as ex:
                    parser.error("Could not parse the date in --incdate parameter. It should be in YYYY-MM-DD format. Please check and try again. \nDate Passed: " + args.incdate + "\n" + str(ex))
            
            #if incperiod (WTD, MTD, YTD, LASTWEEK, LASTMONTH, LASTYEAR) is used instead of incdate
            elif args.incperiod != None:
                if str(args.incperiod).upper() == "WTD":
                    _gv_SampleAPIIncDate = datetime.datetime.strptime(datetime.datetime.now().strftime("%Y-%m-%d"), "%Y-%m-%d")
                    while _gv_SampleAPIIncDate.strftime("%a") != "Sat":
                        _gv_SampleAPIIncDate -= datetime.timedelta(days = 1)
                elif str(args.incperiod).upper() == "MTD":
                    _gv_SampleAPIIncDate = datetime.datetime.strptime(datetime.datetime.now().strftime("%Y-%m") + "-01", "%Y-%m-%d")
                elif str(args.incperiod).upper() == "YTD":
                    _gv_SampleAPIIncDate = datetime.datetime.strptime(datetime.datetime.now().strftime("%Y") + "-01-01", "%Y-%m-%d")
                elif str(args.incperiod).upper() == "LASTWEEK":
                    _gv_SampleAPIIncDate = datetime.datetime.strptime(datetime.datetime.now().strftime("%Y-%m-%d"), "%Y-%m-%d")
                    _gv_SampleAPIIncDate -= datetime.timedelta(days = 7)
                    while _gv_SampleAPIIncDate.strftime("%a") != "Sat":
                        _gv_SampleAPIIncDate -= datetime.timedelta(days = 1)
                elif str(args.incperiod).upper() == "LASTMONTH":
                    _gv_SampleAPIIncDate = datetime.datetime.strptime(datetime.datetime.now().strftime("%Y-%m-%d"), "%Y-%m-%d")
                    _gv_SampleAPIIncDate -= datetime.timedelta(days = 31)
                    while int(_gv_SampleAPIIncDate.strftime("%d")) != 1:
                        _gv_SampleAPIIncDate -= datetime.timedelta(days = 1)
                elif str(args.incperiod).upper() == "LASTYEAR":
                    _gv_SampleAPIIncDate = datetime.datetime.strptime(str(int(datetime.datetime.now().strftime("%Y"))-1) + "-01-01", "%Y-%m-%d")
                else:
                    parser.error("Could not parse the --incperiod parameter. It should be one of WTD, MTD, YTD, LastWeek, LastMonth, LastYear values. \nPlease check and try again. \nPeriod Passed: " + args.incperiod)
                _gv_EmailMessage += "\n Load Type: INC"
                _gv_EmailMessage += "\n Inc Period: " + args.incperiod
                _gv_EmailMessage += "\n Inc Date: " + str(_gv_SampleAPIIncDate)
            else:
                parser.error("Could not parse the --incdate / --incperiod parameter. Please check and try again. \nDate Passed: " + args.incdate + " \nPeriod Passed: " + args.incperiod)
        if not(os.path.isfile(args.output)):
            parser.error("Output File does not exists and --load=Inc is specified")
    
    #if load type is FULL
    elif str(args.load).lower() == "full":
        _gv_SampleAPIIncDate = None
        _gv_EmailMessage += "\n Load Type: FULL"
    else:
        parser.error("Invalid --load parameter passed. Please check and try again. \n--load: " + args.load)

    #Check for Output File
    if not (str(os.path.splitext(args.output)[-1]).lower() == ".hyper" or str(os.path.splitext(args.output)[-1]).lower() == ".tdsx"):
        parser.error("Invalid file specified in --output parameter. File must end in .hyper or .tdsx (existing package file) extension. Please check and try again. \n--output: " + args.output)
    #When output file specified is .tdsx
    elif str(os.path.splitext(args.output)[-1]).lower() == ".tdsx":
        if os.path.isfile(args.output) != True:
            parser.error("Invalid .tdsx Output File specified. Output File does not exists, cannot proceed. --output with .tdsx file requires file to be existing. Please check and try again. \n--output: " + args.output)
        else:
            _gv_OutputPackageFile = args.output
            _gv_ArchiveUnarchiveTempPath += datetime.datetime.now().strftime("_%Y%m%d_%H%M%S")
            _gv_EmailMessage += "\n Output Package File: " + str(_gv_OutputPackageFile)
    #When output file specified is .hyper
    else:
        _gv_OutputHyperFile = args.output
        _gv_EmailMessage += "\n Output Hyper File: " + str(_gv_OutputHyperFile)

    #Check for Package XML file
    if not(args.packagexml == None or (os.path.isfile(args.packagexml) == True and str(os.path.splitext(args.packagexml)[-1]).lower() == ".tds")):
        parser.error("Invalid --packagexml file specified. File is not .tds or does not exists. \n--packagexml: " + args.packagexml)
    else:
        _gv_PackageXMLFile = args.packagexml
        _gv_EmailMessage += "\n Pacakage XML File: " + str(_gv_PackageXMLFile)

    #Check for Backup
    if str(args.backup).lower() == "yes":
        _gv_Backup = True
        if os.path.isfile(args.output) != True:
            parser.error("Invalid Output File specified with --backup = Yes. Output File does not exists, cannot backup. Please check and try again. \n--output: " + args.output)
        if not (os.path.isdir(args.backuppath if args.backuppath != None else ".\\") == True):
            parser.error("Invalid Backup Path specified. Please check and try again. \n--backuppath: " + args.backuppath)
        else:
            _gv_BackupFilePath = args.backuppath
            _gv_EmailMessage += "\n Backup: " + str(_gv_Backup)
            _gv_EmailMessage += "\n Backup File Path: " + str(_gv_BackupFilePath)
    
    #Finished parsing of Arguments
    _gv_EmailMessage += "\n\nFinished Parsing Arguments\n\n"

#Function for writing Log (onscreen or to a file)
def LogFileWriter(LogString):
    global _gv_LogFile, _gv_EmailMessage
    if _gv_LogFile == None:
        if _gv_LogFilePath == None:
            _gv_LogFile = sys.stdout
        else:
            try:
                _gv_LogFile = open(_gv_LogFilePath, "a")
            except Exception as ex:
                print("Error opening Log File. Exiting.\nLog File: " + _gv_LogFile)
                print(str(ex))
                exit(1)
        _gv_LogFile.write("\n\n")
        _gv_LogFile.write("----------------------------------------------------\n")
        _gv_LogFile.write(_gv_ApplicationBanner + "\n")
        _gv_LogFile.write("----------------------------------------------------\n")
        _gv_LogFile.write(datetime.datetime.now().strftime("%m/%d/%y %H:%M:%S") + ": " + "Beginning Script\n")
        _gv_EmailMessage += str(datetime.datetime.now().strftime("%m/%d/%y %H:%M:%S") + ": " + "Beginning Script\n")
    _gv_LogFile.write(datetime.datetime.now().strftime("%m/%d/%y %H:%M:%S") + ": " + LogString + "\n")
    _gv_EmailMessage += str(datetime.datetime.now().strftime("%m/%d/%y %H:%M:%S") + ": " + LogString + "\n")
    _gv_LogFile.flush()

#Function for Sending Email Message
def  SendEmailMessage():
    if _gv_EmailMessage != None and _gv_EmailMessageFrom != None and _gv_EmailMessageTo != None and _gv_EmailServer != None:
        Message = email.message.EmailMessage()
        Message["Subject"] = _gv_ApplicationBanner + " - run of " + str(datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
        Message["From"] = _gv_EmailMessageFrom
        Message["To"] = _gv_EmailMessageTo
        Message.set_content(str(_gv_EmailMessage))
        try:
            with smtplib.SMTP(_gv_EmailServer) as SMTPServer:
                SMTPServer.send_message(Message)
        except Exception as ex:
            LogFileWriter(str(ex))

#Function for Backing up Output File before writing
def FileBackup(OrigPath, NewPath = None):
    if NewPath == None:
        shutil.copyfile(OrigPath, OrigPath + "_" + datetime.datetime.now().strftime("%Y%m%d_%H%M%S"))
        LogFileWriter("File Backed up successfully at location " + OrigPath + "_" + datetime.datetime.now().strftime("%Y%m%d_%H%M%S"))
    else:
        shutil.copy(OrigPath, os.path.join(NewPath, os.path.basename(OrigPath) + "_" + datetime.datetime.now().strftime("%Y%m%d_%H%M%S")))
        LogFileWriter("File Backed up successfully at location " + os.path.join(NewPath, os.path.basename(OrigPath) + "_" + datetime.datetime.now().strftime("%Y%m%d_%H%M%S")))

#Function for Archiving / Unarchiving .tdsx file
def FileArchiveUnarchive(OrigPath, Option):
    global _gv_OutputHyperFile
    
    #Archive File Operation
    if Option == "archive":
        try:
            if str(os.path.splitext(OrigPath)[-1]).lower() == ".tdsx":
                if os.path.isdir(os.path.join(os.path.dirname(OrigPath), _gv_ArchiveUnarchiveTempPath)) == True:
                    os.remove(OrigPath)
                else:
                    raise Exception("Error in archiving .tdsx file. Could not find unarchived extract to re-archive.")
                ZipFile = zipfile.ZipFile(OrigPath, "w", zipfile.ZIP_DEFLATED)
                for DirPath, SubDir, Files in os.walk(os.path.join(os.path.dirname(OrigPath), _gv_ArchiveUnarchiveTempPath)):
                    for FileName in Files:
                        FilePath = os.path.join(DirPath, FileName)
                        if str(os.path.splitext(FileName)[-1]).lower() == ".tds":
                            if _gv_PackageXMLFile == None:
                                ZipFile.write(FilePath, FilePath.replace(os.path.dirname(OrigPath) + "\\" + _gv_ArchiveUnarchiveTempPath + "\\", ""))
                            else:
                                ZipFile.write(_gv_PackageXMLFile, FilePath.replace(os.path.dirname(OrigPath) + "\\" + _gv_ArchiveUnarchiveTempPath + "\\", ""))
                                LogFileWriter("Replaced .tds file: " + FilePath.replace(os.path.dirname(OrigPath) + "\\" + _gv_ArchiveUnarchiveTempPath + "\\", "") + " in the Output File with file --packagexml: " + _gv_PackageXMLFile)
                        else:
                            ZipFile.write(FilePath, FilePath.replace(os.path.dirname(OrigPath) + "\\" + _gv_ArchiveUnarchiveTempPath + "\\", ""))
                ZipFile.close()
                shutil.rmtree(os.path.join(os.path.dirname(OrigPath) , _gv_ArchiveUnarchiveTempPath))
            elif str(os.path.splitext(OrigPath)[-1]).lower() == ".hyper":
                if _gv_PackageXMLFile == None:
                    raise Exception("Error in archiving .tdsx file. Could not find Package XML file to create the archive.")
                ZipFile = zipfile.ZipFile(os.path.join(os.path.dirname(OrigPath), os.path.basename(_gv_PackageXMLFile) + "x"), "w", zipfile.ZIP_DEFLATED)
                ZipFile.write(_gv_OutputHyperFile, _gv_OutputHyperFile.replace(os.path.dirname(_gv_OutputHyperFile) + "\\", ""))
                ZipFile.write(_gv_PackageXMLFile, _gv_PackageXMLFile.replace(os.path.dirname(_gv_PackageXMLFile) + "\\", ""))
                ZipFile.close()
                LogFileWriter("Created new .tdsx Output File " + os.path.join(os.path.dirname(OrigPath), os.path.basename(_gv_PackageXMLFile) + "x") + " based on file --packagexml: " + _gv_PackageXMLFile)
            elif os.path.isdir(OrigPath) == True:
                FilePaths = []
                for DirPath, SubDir, Files in os.walk(OrigPath):
                    for FileName in Files:
                        FilePath = os.path.join(DirPath, FileName)
                        FilePaths.append(FilePath)
                ZipFile = zipfile.ZipFile(os.path.join(OrigPath, os.path.basename(OrigPath) + ".zip"), "w", zipfile.ZIP_DEFLATED)
                for FilePath in FilePaths:
                    ZipFile.write(FilePath, FilePath.replace(os.path.dirname(FilePath) + "\\", ""))
                ZipFile.close()
        except Exception as ex:
            LogFileWriter("Error in archiving file " + OrigPath + ". Please check below error message and try again.")
            LogFileWriter(str(ex))
    
    #Unarchive File Operation
    elif Option == "unarchive":
        try:
            ZipFile = zipfile.ZipFile(OrigPath, "r")
            if str(os.path.splitext(OrigPath)[-1]).lower() == ".tdsx":
                for File_Name in ZipFile.namelist():
                    if str(os.path.splitext(File_Name)[-1]).lower() == ".hyper":
                        _gv_OutputHyperFile = os.path.join(os.path.dirname(OrigPath), _gv_ArchiveUnarchiveTempPath, File_Name)
                        LogFileWriter("Found the Hyper File in the Output Package. Hyper File Name: " + os.path.basename(_gv_OutputHyperFile))
                if _gv_OutputHyperFile == None:
                    raise Exception("Could not find .hyper file in Package File: " + OrigPath)
            ZipFile.extractall(os.path.join(os.path.dirname(OrigPath), _gv_ArchiveUnarchiveTempPath + "\\"))
            ZipFile.close()
        except Exception as ex:
            LogFileWriter("Error in unarchiving file " + OrigPath + ". Please check below error message and try again.")
            LogFileWriter(str(ex))

#Function for controlling max no. of requests per second to API
def SampleAPIConnectionCounter():
    global _gv_SampleAPIReqCounter

    if _gv_SampleAPIMaxRequestsPerSec != 0 and _gv_SampleAPIMaxRequestsPerSec != None:
        if _gv_SampleAPIReqCounter["LastAccessed"] == datetime.datetime.now().strftime("%m/%d/%y %H:%M:%S"):
            if _gv_SampleAPIReqCounter["RequestCount"] >= _gv_SampleAPIMaxRequestsPerSec - 1:
                time.sleep(1)
                SampleAPIConnectionCounter()
            else:
                _gv_SampleAPIReqCounter["RequestCount"] += 1
        else:
            _gv_SampleAPIReqCounter["LastAccessed"] = datetime.datetime.now().strftime("%m/%d/%y %H:%M:%S")
            _gv_SampleAPIReqCounter["RequestCount"] = 1

##Function for Initial Connection to Sample API
def SampleAPIInitialConnection (SampleAPIURL, SampleAPIKey):
    global _gv_SampleAPIURL, _gv_SampleAPIAuthorization, _gv_SampleAPIHeader, _gv_SampleAPIBaseURIS, _gv_SampleAPIResponse
    _gv_SampleAPIURL = SampleAPIURL
    _gv_SampleAPIAuthorization = "Basic " + str(base64.b64encode(SampleAPIKey.encode("utf-8")), "utf-8")
    _gv_SampleAPIHeader = {"Authorization":_gv_SampleAPIAuthorization, "Content-Type":_gv_SampleAPIContentType}
    _gv_SampleAPIBaseURIS = {"users_base_url":_gv_SampleAPIURL + "/users", "resource_base_url":_gv_SampleAPIURL + "/unknown"}
    
    ##Initiating Connection
    LogFileWriter("Accessing Base Users API")
    SampleAPIConnectionCounter()
    _gv_SampleAPIResponse = requests.request("GET", _gv_SampleAPIBaseURIS["users_base_url"], headers = _gv_SampleAPIHeader, timeout = _gv_SampleAPITimeOut, allow_redirects = False)
    if _gv_SampleAPIResponse.status_code == 200:
        LogFileWriter("Base Users API Retrieved")
        _gv_SampleAPICookies = _gv_SampleAPIResponse.cookies
        SampleAPIJSONS = []
        SampleAPIJSON = _gv_SampleAPIResponse.json()
        SampleAPIJSONS.append(SampleAPIJSON)
        ##Retreive all pages if API uses pagination
        while SampleAPIJSON.get("page") != SampleAPIJSON.get("total_pages"):
            _gv_SampleAPIResponse = requests.request("GET", _gv_SampleAPIBaseURIS["users_base_url"] + "?page=" + str(int(SampleAPIJSON.get("page")) + 1), headers = _gv_SampleAPIHeader, timeout = _gv_SampleAPITimeOut, allow_redirects = False, cookies=_gv_SampleAPICookies)
            SampleAPIJSON = _gv_SampleAPIResponse.json()
            SampleAPIJSONS.append(SampleAPIJSON)
        for SampleAPIJSON in SampleAPIJSONS:
            if len(SampleAPIJSON["data"]) <= 0:
                LogFileWriter("No Users found in Sample API for data retreival")
                LogFileWriter("Exiting Script")
                raise Exception("No Users found in Sample API for data retreival")
    else:
        LogFileWriter("Error in receiving Base Users API")
        LogFileWriter("Please check connection parameters again.")
        LogFileWriter("Exiting Script")
        raise Exception("Error Establishing Connection, Unable to get Base Users API \nSample API URL: " + SampleAPIURL + " \nSample API Key: " + SampleAPIKey + " \nSample API Authorization: " + _gv_SampleAPIAuthorization + " \nSample API Base Users URL: " + _gv_SampleAPIBaseURIS["users_base_url"])

##Function for iterating through next pages of API if API uses pagination
def SampleAPIGetDataNextPage(SampleAPIURL, CurrentPageNumber = None):
    global _gv_SampleAPIResponse
    SampleAPIConnectionCounter()
    _gv_SampleAPIResponse = requests.request("GET", SampleAPIURL + ("&" if str(SampleAPIURL).find("?") > 0 else "?") + "page=" + str(int(CurrentPageNumber if CurrentPageNumber != None else 0) + 1), headers = _gv_SampleAPIHeader, timeout = _gv_SampleAPITimeOut, allow_redirects = False, cookies=_gv_SampleAPICookies)
    if _gv_SampleAPIResponse.status_code == 200:
        SampleAPIJSON = _gv_SampleAPIResponse.json()
        return SampleAPIJSON
    else:
        LogFileWriter("Error in receiving data from Sample API")
        LogFileWriter("Please check log and try again.")
        LogFileWriter("Exiting Script")
        raise Exception("Error in receiving data from Sample API, Unable to get data from URL \nSample API URL: " + SampleAPIURL + ("&" if str(SampleAPIURL).find("?") > 0 else "?") + "page=" + (int(CurrentPageNumber if CurrentPageNumber != None else 0) + 1))

##Function for getting data from the API
def SampleAPIGetData(SampleAPIURL, SampleAPIURLPARAMS = None):
    global _gv_SampleAPIResponse
    SampleAPIConnectionCounter()
    _gv_SampleAPIResponse = requests.request("GET", SampleAPIURL + (SampleAPIURLPARAMS if SampleAPIURLPARAMS != None else ""), headers = _gv_SampleAPIHeader, timeout = _gv_SampleAPITimeOut, allow_redirects = False, cookies=_gv_SampleAPICookies)
    if _gv_SampleAPIResponse.status_code == 200:
        SampleAPIJSON = _gv_SampleAPIResponse.json()
        SampleAPIJSONS = []
        SampleAPIJSONS.append(SampleAPIJSON)
        ExceptionCount = 0
        while SampleAPIJSON.get("page") != SampleAPIJSON.get("total_pages"):
            try:
                SampleAPIJSON = SampleAPIGetDataNextPage(SampleAPIURL + (SampleAPIURLPARAMS if SampleAPIURLPARAMS != None else ""), SampleAPIJSON.get("page"))
                SampleAPIJSONS.append(SampleAPIJSON)
                ExceptionCount = 0
            except Exception as ex:
                ExceptionCount += 1
                if ExceptionCount <=5:
                    LogFileWriter("Warning: " + str(ex))
                else:
                    LogFileWriter("Too Many Exceptions: " + str(ex))
                    raise Exception("Too Many Exceptions: " + str(ex))
        return SampleAPIJSONS
    else:
        LogFileWriter("Error in receiving data from Sample API")
        LogFileWriter("Please check log and try again.")
        LogFileWriter("Exiting Script")
        raise Exception("Error in receiving data from Sample API, Unable to get data from URL \nSample API URL: " + SampleAPIURL)

#Function for controlling Tableau Hyper Process
def TabHyperProcessControl(ControlOption, HyperFilePath = None):
    global _gv_TabHyperProcess, _gv_TabHyperCon

    #Tableau Hyper Process start operation
    if ControlOption == "start":
        try:
            try:
                _gv_TabHyperProcess = tableauhyperapi.HyperProcess(_gv_TabHyperSendUsageStats)
            except tableauhyperapi.UnclosedObjectWarning as ex:
                pass
            if _gv_TabHyperProcess.is_open == True:
                LogFileWriter("Started Tableau Hyper Process")
            else:
                LogFileWriter("Tableau Hyper Process not started")
                raise Exception("Tableau Hyper Process not started")
            try:
                _gv_TabHyperCon = tableauhyperapi.Connection(_gv_TabHyperProcess.endpoint, HyperFilePath, tableauhyperapi.CreateMode.CREATE_IF_NOT_EXISTS)
            except tableauhyperapi.UnclosedObjectWarning as ex:
                pass
            if _gv_TabHyperCon.is_open == True:
                LogFileWriter("Opened Connection to Tableau Hyper File: \n" + HyperFilePath)
            else:
                LogFileWriter("Could not open Connection to Tableau Hyper File: " + HyperFilePath + " . Please check path and try again.")
                raise Exception("Could not open Connection to Tableau Hyper File: " + HyperFilePath + " . Please check path and try again.")
        except tableauhyperapi.HyperException as ex:
            LogFileWriter("There is exception in starting Tableau Hyper Process. Exiting...")
            LogFileWriter(str(ex))
            raise Exception("There is exception in starting Tableau Hyper Process.")

    #Tableau Hyper Process shutdown operation    
    elif ControlOption == "shutdown":
        try:
            _gv_TabHyperCon.close()
            _gv_TabHyperProcess.shutdown()
            if _gv_TabHyperProcess.is_open == False:
                    LogFileWriter("Stopped Tableau Hyper Process")
            else:
                _gv_TabHyperProcess.close()
                LogFileWriter("Forcefully stopped Tableau Hyper Process")
            _gv_TabHyperProcess = None
        except tableauhyperapi.HyperException as ex:
            LogFileWriter("There is exception in stopping Tableau Hyper Process. Forcefully Stopping it...")
            LogFileWriter(str(ex))
            _gv_TabHyperProcess.close()
            LogFileWriter("Forcefully stopped Tableau Hyper Process")
    
    #Invalid option passed to the Function
    else:
        LogFileWriter("Invalid Control Option Passed to Hyper Process Control: " + ControlOption)
        raise Exception("Invalid Control Option Passed to Hyper Process Control: " + ControlOption)

#Function for wrting Extract Log Table inside Tableau Hyper File
def TabHyperExtractLogWriter(Table_Affected, Action, Rows_Affected, Comment = None):
    global _gv_TabHyperExtractLogTable
    if _gv_TabHyperExtractLogTable != None:
        Table_Name_Log = tableauhyperapi.TableName(_gv_TabHyperExtractLogTable)
        Column_Definition_Log = [
            tableauhyperapi.TableDefinition.Column("insert_timestamp", tableauhyperapi.SqlType.timestamp()),
            tableauhyperapi.TableDefinition.Column("table_name", tableauhyperapi.SqlType.varchar(100)),
            tableauhyperapi.TableDefinition.Column("action_type", tableauhyperapi.SqlType.varchar(100)),
            tableauhyperapi.TableDefinition.Column("rows_affected", tableauhyperapi.SqlType.int()),
            tableauhyperapi.TableDefinition.Column("comments", tableauhyperapi.SqlType.text())
        ]
        Table_Definition_Log = tableauhyperapi.TableDefinition(Table_Name_Log, Column_Definition_Log, tableauhyperapi.Persistence.PERMANENT)

        _gv_TabHyperCon.catalog.create_table_if_not_exists(Table_Definition_Log)

        with tableauhyperapi.Inserter(_gv_TabHyperCon, Table_Definition_Log) as Log_Inserter:
            try:
                Log_Inserter.add_row([datetime.datetime.now(), Table_Affected, Action, Rows_Affected, Comment])
                Log_Inserter.execute()
            except tableauhyperapi.HyperException as ex:
                LogFileWriter("Error writing Hyper Extract Log Table. Would not try to write again. \n Extract Log Table: " + _gv_TabHyperExtractLogTable)
                LogFileWriter(str(ex))
                _gv_TabHyperExtractLogTable = None

#Function for re-creating Output Hyper File in case of FULL Load
def TabSampleAPIClearData():
    if _gv_SampleAPIIncDate == None:
        try:
            _gv_TabHyperCon.catalog.detach_all_databases()
            _gv_TabHyperCon.catalog.drop_database_if_exists(_gv_OutputHyperFile)
            _gv_TabHyperCon.catalog.create_database_if_not_exists(_gv_OutputHyperFile)
            _gv_TabHyperCon.catalog.attach_database(_gv_OutputHyperFile)
        except tableauhyperapi.HyperException as ex:
            LogFileWriter("Error occurred while truncating Hyper File for Full Load. Please check below error message and try again.")
            LogFileWriter(str(ex))
            raise Exception("Error occurred while truncating Hyper File for Full Load")

#Function for writing the data to Output Hyper File
def TabSampleAPIWriteData(WriteDataSet, Extract_Comment = None):
    
    #Define Tables for Load and Stage
    Table_Name_Load = tableauhyperapi.TableName(_gv_TabHyperLoadSchema, WriteDataSet["table_name"])
    Table_Name_Stage = tableauhyperapi.TableName(WriteDataSet["table_name"] + _gv_TabHyperStageTablesSuffix)
    Table_Definition_Load = None
    Table_Definition_Stage = None
    Extract_Comment = " for " + Extract_Comment if Extract_Comment != None else ""

    #Define Columns for the Tables based on Write Data Set
    Column_Definition = []
    for index in range(0, len(WriteDataSet["columns"])):
        Column_Definition.append(tableauhyperapi.TableDefinition.Column(WriteDataSet["columns"][index], WriteDataSet["columns_tabdatatypes"][index]))

    #Create Schema if it does not exists
    _gv_TabHyperCon.catalog.create_schema_if_not_exists(_gv_TabHyperLoadSchema)
    
    #Create Load Table if it does not exists
    if not(_gv_TabHyperCon.catalog.has_table(Table_Name_Load)):
        Table_Definition_Load = tableauhyperapi.TableDefinition(Table_Name_Load, Column_Definition, tableauhyperapi.Persistence.PERMANENT)
        _gv_TabHyperCon.catalog.create_table_if_not_exists(Table_Definition_Load)
        TabHyperExtractLogWriter(str(Table_Name_Load.name.unescaped), "CREATED", 0, "Table Created")
    else:
        Table_Definition_Load = _gv_TabHyperCon.catalog.get_table_definition(Table_Name_Load)
        Check_Columns_Load = [str(Column.name.unescaped) for Column in Table_Definition_Load.columns]
        Check_Columns_Load.sort()
        Check_Columns_DataSet = WriteDataSet["columns"].copy()
        Check_Columns_DataSet.sort()
        if not(Check_Columns_Load == Check_Columns_DataSet):
            LogFileWriter("Columns for Table: " + WriteDataSet["table_name"] + " is not same as in Hyper file. Please check below:")
            LogFileWriter("Hyper File Columns: " + str([str(Column.name.unescaped) for Column in Table_Definition_Load.columns]))
            LogFileWriter("Columns from SampleAPI: " + str(WriteDataSet["columns"]))
            raise Exception("Mismatched column between Hyper File and Sample API Data Schema for table: " + WriteDataSet["table_name"])
    
    #Create Stage Table if it does not exists
    if not(_gv_TabHyperCon.catalog.has_table(Table_Name_Stage)):
        Table_Definition_Stage = tableauhyperapi.TableDefinition(Table_Name_Stage, Column_Definition, tableauhyperapi.Persistence.TEMPORARY if not _gv_TabHyperStageUsePersistentTables else tableauhyperapi.Persistence.PERMANENT)
        _gv_TabHyperCon.catalog.create_table_if_not_exists(Table_Definition_Stage)
        TabHyperExtractLogWriter(str(Table_Name_Stage.name.unescaped), "CREATED", 0, "Table Created")
    else:
        Table_Definition_Stage = _gv_TabHyperCon.catalog.get_table_definition(Table_Name_Stage)

    #Execute Pre SQL Query
    if len(WriteDataSet["pre_sql_load"]) != 0:
        for Execute_Command in WriteDataSet["pre_sql_load"]:
            try:
                Execute_Result = _gv_TabHyperCon.execute_command(str(Execute_Command).replace("__table_name__", str(Table_Name_Load)))
                TabHyperExtractLogWriter(str(Table_Name_Load.name.unescaped), "PRE_SQL_LOAD", Execute_Result, "Executed Pre SQL " + str(WriteDataSet["pre_sql_load"].index(Execute_Command) + 1) + Extract_Comment)
            except tableauhyperapi.HyperException as ex:
                LogFileWriter("There are errors in Pre SQL Load for table " + WriteDataSet["table_name"] + ".\n Error in Pre SQL: \n" + str(Execute_Command) + "\nPlease check below for more details.")
                LogFileWriter(str(ex))
                TabHyperProcessControl("shutdown")
                raise Exception("Errors in writing Hyper File. Please check errors and try again. Exiting...")
    
    #Truncate Stage Table
    Execute_Command = f"TRUNCATE TABLE {Table_Name_Stage}"
    try:
        Execute_Result = _gv_TabHyperCon.execute_command(Execute_Command)
        TabHyperExtractLogWriter(str(Table_Name_Stage.name.unescaped), "TRUNCATE", Execute_Result, "Stage Table Truncated" + Extract_Comment)
    except tableauhyperapi.HyperException as ex:
        LogFileWriter("There are errors in truncating Hyper File Stage Table for: " + WriteDataSet["table_name"] + ". Please check below for more details.")
        LogFileWriter(str(ex))
        TabHyperProcessControl("shutdown")
        raise Exception("Errors in writing Hyper File. Please check errors and try again. Exiting...")
    
    #Insert Data into Stage Table
    with tableauhyperapi.Inserter(_gv_TabHyperCon, Table_Definition_Stage, WriteDataSet["columns"]) as TabInserter:
        try:
            TabInserter.add_rows(WriteDataSet["rows"])
            TabInserter.execute()
            TabHyperExtractLogWriter(str(Table_Name_Stage.name.unescaped), "INSERT", len(WriteDataSet["rows"]), "Rows Inserted" + Extract_Comment)
        except tableauhyperapi.HyperException as ex:
            LogFileWriter("There are errors in writing data to Hyper File Table: " + WriteDataSet["table_name"] + " in stage. Please check below for more details.")
            LogFileWriter(str(ex))
            TabHyperProcessControl("shutdown")
            raise Exception("Errors in writing Hyper File. Please check errors and try again. Exiting...")

    #Merge Insert data Load Table
    Execute_Command = f"INSERT INTO {Table_Name_Load} (" + ", ".join(WriteDataSet["columns"])
    Execute_Command += f") SELECT " + ", ".join(WriteDataSet["columns"]) + f" FROM {Table_Name_Stage}"
    try:
        Execute_Result = _gv_TabHyperCon.execute_command(Execute_Command)
        TabHyperExtractLogWriter(str(Table_Name_Load.name.unescaped), "MERGE_INSERT", Execute_Result, "Merge Load with Stage Table Step 1" + Extract_Comment)
    except tableauhyperapi.HyperException as ex:
        LogFileWriter("There are errors in insert merging data to Hyper File Table: " + WriteDataSet["table_name"] + ". Please check below for more details.")
        LogFileWriter(str(ex))
        TabHyperProcessControl("shutdown")
        raise Exception("Errors in writing Hyper File. Please check errors and try again. Exiting...")

    #Merge Delete data from Load Table
    Execute_Command = f"DELETE FROM {Table_Name_Load} DELQ USING (SELECT " 
    Execute_Command += f", ".join(WriteDataSet["primary_column"]) + f", MAX(" + WriteDataSet["insert_date_key"] + f") AS insert_date_key FROM {Table_Name_Load} "
    Execute_Command += f"GROUP BY " + f", ".join(WriteDataSet["primary_column"]) + f") MAPQ " 
    Execute_Command += f"WHERE DELQ." + f" || {tableauhyperapi.escape_string_literal('-')} || DELQ.".join(WriteDataSet["primary_column"]) 
    Execute_Command += f" = MAPQ." + f" || {tableauhyperapi.escape_string_literal('-')} || MAPQ.".join(WriteDataSet["primary_column"]) + f" AND DELQ." + WriteDataSet["insert_date_key"] + f" <> MAPQ.insert_date_key"
    try:
        Execute_Result = _gv_TabHyperCon.execute_command(Execute_Command)
        TabHyperExtractLogWriter(str(Table_Name_Load.name.unescaped), "MERGE_DELETE", Execute_Result, "Merge Load with Stage Table Step 2" + Extract_Comment)
    except tableauhyperapi.HyperException as ex:
        LogFileWriter("There are errors in delete merging data from Hyper File Table: " + WriteDataSet["table_name"] + ". Please check below for more details.")
        LogFileWriter(str(ex))
        TabHyperProcessControl("shutdown")
        raise Exception("Errors in writing Hyper File. Please check errors and try again. Exiting...")

    #Execute Post SQL Query
    if len(WriteDataSet["post_sql_load"]) != 0:
        for Execute_Command in WriteDataSet["post_sql_load"]:
            try:
                Execute_Result = _gv_TabHyperCon.execute_command(str(Execute_Command).replace("__table_name__", str(Table_Name_Load)))
                TabHyperExtractLogWriter(str(Table_Name_Load.name.unescaped), "POST_SQL_LOAD", Execute_Result, "Executed Post SQL " + str(WriteDataSet["post_sql_load"].index(Execute_Command) + 1) + Extract_Comment)
            except tableauhyperapi.HyperException as ex:
                LogFileWriter("There are errors in Post SQL Load for table " + WriteDataSet["table_name"] + ".\n Error in Post SQL: \n" + str(Execute_Command) + "\nPlease check below for more details.")
                LogFileWriter(str(ex))
                TabHyperProcessControl("shutdown")
                raise Exception("Errors in writing Hyper File. Please check errors and try again. Exiting...")

##Function to get and parse data from Sample API
def SampleAPIParseData(SampleAPIObjectType = None):
    
    #Start Tableau Hyper Process
    try:
        LogFileWriter("Starting Tableau Hyper Process")
        TabHyperProcessControl("start", _gv_OutputHyperFile)
    except Exception as ex:
        LogFileWriter("Error Occurred while starting Tableau Hyper Process. See below error messages for more details.")
        LogFileWriter(str(ex))
        SendEmailMessage()
        _gv_LogFile.close()
        exit(1)
    
    ##Get Data for the different API Modules from Sample API
    try:
        if SampleAPIObjectType == None:
            LogFileWriter("Starting to Parse Sample API Data for All")
            TabSampleAPIClearData()
        if SampleAPIObjectType == "users" or SampleAPIObjectType == None:
            LogFileWriter("Starting to Parse Sample API Data for Users")
            SampleAPIJSONS = SampleAPIGetData(_gv_SampleAPIURL + "/users", ("?start_date=" + _gv_SampleAPIIncDate if _gv_SampleAPIIncDate != None else None))
            TabSampleAPIWriteData(SampleDataSchema.SampleAPIDecoder(SampleAPIJSONS, "users"))
            LogFileWriter("Finished Parsing Sample API Data for Users")
        if SampleAPIObjectType == "resource" or SampleAPIObjectType == None:
            LogFileWriter("Starting to Parse Sample API Data for Resource")
            SampleAPIJSONS = SampleAPIGetData(_gv_SampleAPIURL + "/unknown", ("?start_date=" + _gv_SampleAPIIncDate if _gv_SampleAPIIncDate != None else None))
            TabSampleAPIWriteData(SampleDataSchema.SampleAPIDecoder(SampleAPIJSONS, "resource"))
            LogFileWriter("Finished Parsing Sample API Data for Resource")
        if SampleAPIObjectType == None:
            LogFileWriter("Finished Parsing Sample API Data for All")
    except Exception as ex:
        LogFileWriter("Error parsing Sample API data. Please check below error and try again.")
        LogFileWriter(str(ex))
        raise Exception ("Error parsing Sample API data.")

    #Stop Tableau Hyper Process
    finally:
        try:
            LogFileWriter("Stopping Tableau Hyper Process")
            TabHyperProcessControl("shutdown")
        except Exception as ex:
            LogFileWriter("Error Occurred while stopping Tableau Hyper Process. See below error messages for more details.")
            LogFileWriter(str(ex))
            SendEmailMessage()
            _gv_LogFile.close()
            exit(1)


#Main Routine, Execution starts here
if __name__ == "__main__":
    
    #1 - Parse Arguments
    try:
        ParseArguments()
    except Exception as ex:
        print("Error Occured while parsing input parameters. Please check the parameters and try again.")
        print(str(ex))
        exit(1)

    #2 - Backup existing Output File
    try:
        if _gv_Backup == True:
            FileBackup((_gv_OutputHyperFile if _gv_OutputHyperFile != None else _gv_OutputPackageFile), _gv_BackupFilePath)
            LogFileWriter("Pre-Existing Output File Backup Done")
    except Exception as ex:
        LogFileWriter("Error Occurred while backing up Pre-Existing Output file. Please check paths and try again.")
        LogFileWriter(str(ex))
        print("Error Occurred while backing up Pre-Existing Output file. Please check paths and try again.")
        print(ex)
        SendEmailMessage()
        _gv_LogFile.close()
        exit(1)

    #3 - Unarchive .tdsx File
    try:
        if _gv_OutputHyperFile == None:
            FileArchiveUnarchive(_gv_OutputPackageFile, "unarchive")
            LogFileWriter("Unarchived Pre-Existing Output File")
    except Exception as ex:
        LogFileWriter("Error Occurred while unarchiving Pre-Existing Output file. Please check paths and try again.")
        LogFileWriter(str(ex))
        print("Error Occurred while unarchiving Pre-Existing Output file. Please check paths and try again.")
        print(ex)
        SendEmailMessage()
        _gv_LogFile.close()
        exit(1)

    #4 - Perform Initial Connection
    try:
        LogFileWriter("Initiating Connection")
        SampleAPIInitialConnection(_gv_SampleAPIURL, _gv_SampleAPIKey)
    except Exception as ex:
        LogFileWriter("Error Occurred while initiating connection. Please check connection parameters and try again.")
        LogFileWriter(str(ex))
        print("Error Occurred while initiating connection. Please check connection parameters and try again.")
        print(ex)
        SendEmailMessage()
        _gv_LogFile.close()
        exit(1)

    #5 - Get and Parse Data from Sample API, write to Output Hyper File
    try:
        LogFileWriter("Starting to parse Sample API Data")
        SampleAPIParseData()
    except Exception as ex:
        LogFileWriter("Error Occurred while parsing Sample API Data. See below error messages for more details.")
        LogFileWriter(str(ex))
        SendEmailMessage()
        _gv_LogFile.close()
        exit(1)

    #6 - Archive .tdsx File
    try:
        if not (_gv_OutputPackageFile == None and _gv_PackageXMLFile == None):
            FileArchiveUnarchive((_gv_OutputPackageFile if _gv_OutputPackageFile != None else _gv_OutputHyperFile), "archive")
            LogFileWriter("Archived Pre-Existing Output File")
    except Exception as ex:
        LogFileWriter("Error Occurred while archiving Pre-Existing Output file. Please check paths and try again.")
        LogFileWriter(str(ex))
        print("Error Occurred while archiving Pre-Existing Output file. Please check paths and try again.")
        print(ex)
        SendEmailMessage()
        _gv_LogFile.close()
        exit(1)
    
    #End of Execution, Send Email and Close Log File
    LogFileWriter("Finished running the Script.\n")
    print("\nFinished running the Script.\nPlease check " + ("Log File: " + _gv_LogFilePath if _gv_LogFilePath != None else "above screen messages") + (" and Extract Log Table: " + _gv_TabHyperExtractLogTable + " in the Output file" if _gv_TabHyperExtractLogTable != None else "") + " for more details.\n")
    SendEmailMessage()
    _gv_LogFile.close()
    


Now lot of things are happening in this file. The best way to analyze is by looking at the __main__ routine in the very end of the script (line 652)


What we are doing here is
  1. Parsing the input arguments
  2. Backing up the existing output file
  3. Unarchiving the .tdsx file if it was specified as output
  4. Performing Initial Connection to Cloud API
  5. Thereafter getting the data from Cloud API and parsing it, to store it into .hyper file format
  6. Archiving the .hyper file into .tdsx file if it was specified as output
  7. Sending Email and completing Execution

Each of these steps above make individual function call to the code above. The output log generated is based on the processing of same steps as well. Explaining all of these functions and steps in detail would be tedious task and would be impractical so it is out of scope for this blog .

The code includes comment in most places to help you walk through, but here below are the most important functions which you need to modify to make them work with your API.
  • Modify the Variables after Global Variables in the top part of script at Line 37 to define basic inputs which would stay constant across script runs.
  • Modify Input Arguments as required in ParseArguments function at Line 74 to match to your required inputs.
  • Modify Initial Connection Function SampleAPIInitialConnection at Line 319 to have the routine for initial connection to your  Cloud API (could be any simple test).
  • Modify logic for getting next pages at Line 337 and in SampleAPIGetDataNextPage function at Line 353 if your API uses pagination.
  • Modify URL format and other parameters sent in the actual request to the API in SampleAPIGetData function at Line 367.
  • This is a big one and most important - Add and Update If conditions in SampleAPIParseData function at Line 603 to conditionally send requests for different API modules and call equivalent parsing functions from SampleDataSchema.py file imported at Line 22.

--> Let's look at SampleDataSchema.py

Warning! This looks simple here from Sample API but takes the most time when doing real world requirement, in defining each field for the JSON received from the API!!

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
#Basic Imports
import datetime

#Import for defining data types for Output Hyper File
from tableauhyperapi import SqlType

#Function for Base Data Set Definition
def SampleAPIDataSetBaseDef():
    return {"table_name":"", "columns":[], "columns_tabdatatypes": [], "primary_column": [], "insert_date_key": "insert_date" , "pre_sql_load": [], "post_sql_load": [], "rows": []}

##Function for identifying decoding function to be used based on Object Type passed
def SampleAPIDecoder(SampleAPIJSONS, SampleAPIObjectType):
    if SampleAPIObjectType == "users":
        return SampleAPIDecoderUsers(SampleAPIJSONS)
    if SampleAPIObjectType == "resource":
        return SampleAPIDecoderResource(SampleAPIJSONS)

##Decoding Function - Users
def SampleAPIDecoderUsers(SampleAPIJSONS):
    DataSet = SampleAPIDataSetBaseDef()
    DataSet["table_name"] = "Users"
    DataSet["primary_column"].append("user_id")
    Insert_Date = datetime.datetime.now()

    #Define Columns
    DataSet["columns"].append("user_id")
    DataSet["columns"].append("first_name")
    DataSet["columns"].append("last_name")
    DataSet["columns"].append("email")
    DataSet["columns"].append("avatar_url")
    DataSet["columns"].append("insert_date")

    #Define Column Tableau Data Types in same order as Columns above 
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(20))    #"user_id"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(50))    #"first_name"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(50))    #"last_name"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(100))   #"email"
    DataSet["columns_tabdatatypes"].append(SqlType.text())         #"avatar_url"
    DataSet["columns_tabdatatypes"].append(SqlType.timestamp())    #"insert_date"

    #Define Data rows w.r.t Data parsed from SampleAPIJSONS in same order as Columns above 
    for SampleAPIJSON in SampleAPIJSONS:
        for SampleAPIData in SampleAPIJSON["data"]:
            Row = []
            Row.append(str(SampleAPIData["id"]))      #"user_id"
            Row.append(SampleAPIData.get("first_name"))                 #"first_name"
            Row.append(SampleAPIData.get("last_name"))                 #"last_name"
            Row.append(SampleAPIData.get("email"))        #"email"
            Row.append(SampleAPIData.get("avatar"))                     #"avatar_url"
            Row.append(Insert_Date)                               #"insert_date"
            DataSet["rows"].append(Row)

    #Post SQL
    Execute_Command = "DELETE FROM __table_name__ DELQ USING " 
    Execute_Command += "(SELECT MAX(ctid) as ctid, user_id, insert_date FROM __table_name__ GROUP BY user_id, insert_date) MAPQ "
    Execute_Command += "WHERE DELQ.user_id = MAPQ.user_id AND DELQ.insert_date = MAPQ.insert_date AND DELQ.ctid <> MAPQ.ctid"
    DataSet["post_sql_load"].append(Execute_Command)

    #Return the DataSet
    return DataSet

##Decoding Function - Resource
def SampleAPIDecoderResource(SampleAPIJSONS):
    DataSet = SampleAPIDataSetBaseDef()
    DataSet["table_name"] = "Resource"
    DataSet["primary_column"].append("resource_id")
    Insert_Date = datetime.datetime.now()

    #Define Columns
    DataSet["columns"].append("resource_id")
    DataSet["columns"].append("resource_name")
    DataSet["columns"].append("year")
    DataSet["columns"].append("color")
    DataSet["columns"].append("pantone_value")
    DataSet["columns"].append("insert_date")

    #Define Column Tableau Data Types in same order as Columns above 
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(20))    #"resource_id"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(50))    #"resource_name"
    DataSet["columns_tabdatatypes"].append(SqlType.int())          #"year"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(20))    #"color"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(20))    #"pantone_value"
    DataSet["columns_tabdatatypes"].append(SqlType.timestamp())    #"insert_date"

    #Define Data rows w.r.t Data parsed from SampleAPIJSONS in same order as Columns above 
    for SampleAPIJSON in SampleAPIJSONS:
        for SampleAPIData in SampleAPIJSON["data"]:
            Row = []
            Row.append(str(SampleAPIData["id"]))      #"resource_id"
            Row.append(SampleAPIData.get("name"))                     #"resource_name"
            Row.append(SampleAPIData.get("year"))                     #"year"
            Row.append(SampleAPIData.get("color"))        #"color"
            Row.append(SampleAPIData.get("pantone_value"))              #"pantone_value"
            Row.append(Insert_Date)                               #"insert_date"
            DataSet["rows"].append(Row)

    #Return the DataSet
    return DataSet


Now in here you are defining the base data structure for an object in the form of a Python Dictionary which would become a table in Tableau, defining a Decoder function which would be entry point for calling actual parsing functions from Main.py and then defining the actual parsing functions.

What should be really concerning here is:
  • Defining the entry points for parsing functions in the Decoder function, so this could be called from the Main.py script
  • Defining the actual parsing functions. These has below areas to be considered.
    • Provide Table Name and define Primary Column(s) for the going to be table(s).
    • Define the various Column Names first
    • Define the Data Type for these Columns in the same order as above
    • Define how to parse the actual JSON object and assign values to the Columns in the same order as above
    • Define any Pre/Post SQL(s) if required for processing before/after the data load happens for the table.
  • You could also use Faker library here as given in SampleDataSchema_Fake.py script to randomize data input for various Columns if needed.

--> SampleDataSchema_Fake.py code

Warning! No Warnings here!!

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
#Basic Imports
import datetime

#Import for defining data types for Output Hyper File
from tableauhyperapi import SqlType

#Import Faker for generating random data
from faker import Faker
from faker.providers import address
from faker.providers import company
from faker.providers import internet
from faker.providers import person
from faker.providers import phone_number
from faker.providers import profile
from faker.providers import lorem


#Function for Base Data Set Definition
def SampleAPIDataSetBaseDef():
    return {"table_name":"", "columns":[], "columns_tabdatatypes": [], "primary_column": [], "insert_date_key": "insert_date" , "pre_sql_load": [], "post_sql_load": [], "rows": []}

##Function for identifying decoding function to be used based on Object Type passed
def SampleAPIDecoder(SampleAPIJSONS, SampleAPIObjectType):
    if SampleAPIObjectType == "users":
        return SampleAPIDecoderUsers(SampleAPIJSONS)
    if SampleAPIObjectType == "resource":
        return SampleAPIDecoderResource(SampleAPIJSONS)

##Decoding Function - Users
def SampleAPIDecoderUsers(SampleAPIJSONS):
    DataSet = SampleAPIDataSetBaseDef()
    DataSet["table_name"] = "Users"
    DataSet["primary_column"].append("user_id")
    Insert_Date = datetime.datetime.now()

    #Define Columns
    DataSet["columns"].append("user_id")
    DataSet["columns"].append("first_name")
    DataSet["columns"].append("last_name")
    DataSet["columns"].append("email")
    DataSet["columns"].append("avatar_url")
    DataSet["columns"].append("insert_date")

    #Define Column Tableau Data Types in same order as Columns above 
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(20))    #"user_id"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(50))    #"first_name"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(50))    #"last_name"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(100))   #"email"
    DataSet["columns_tabdatatypes"].append(SqlType.text())         #"avatar_url"
    DataSet["columns_tabdatatypes"].append(SqlType.timestamp())    #"insert_date"

    fake = Faker()
    fake.add_provider(address)
    fake.add_provider(company)
    fake.add_provider(internet)
    fake.add_provider(person)
    fake.add_provider(phone_number)
    fake.add_provider(profile)

    #Define Data rows w.r.t Data parsed from SampleAPIJSONS in same order as Columns above 
    for SampleAPIJSON in SampleAPIJSONS:
        for SampleAPIData in SampleAPIJSON["data"]:
            Row = []
            Row.append(str(SampleAPIData["id"]))      #"user_id"
            Row.append(fake.first_name())                            #"first_name"
            Row.append(fake.last_name())                             #"last_name"
            Row.append(fake.email())                        #"email"
            Row.append(fake.uri())                                      #"avatar_url"
            Row.append(Insert_Date)                               #"insert_date"
            DataSet["rows"].append(Row)

    #Post SQL
    Execute_Command = "DELETE FROM __table_name__ DELQ USING " 
    Execute_Command += "(SELECT MAX(ctid) as ctid, user_id, insert_date FROM __table_name__ GROUP BY user_id, insert_date) MAPQ "
    Execute_Command += "WHERE DELQ.user_id = MAPQ.user_id AND DELQ.insert_date = MAPQ.insert_date AND DELQ.ctid <> MAPQ.ctid"
    DataSet["post_sql_load"].append(Execute_Command)

    #Return the DataSet
    return DataSet

##Decoding Function - Resource
def SampleAPIDecoderResource(SampleAPIJSONS):
    DataSet = SampleAPIDataSetBaseDef()
    DataSet["table_name"] = "Resource"
    DataSet["primary_column"].append("resource_id")
    Insert_Date = datetime.datetime.now()

    #Define Columns
    DataSet["columns"].append("resource_id")
    DataSet["columns"].append("resource_name")
    DataSet["columns"].append("year")
    DataSet["columns"].append("color")
    DataSet["columns"].append("pantone_value")
    DataSet["columns"].append("insert_date")

    #Define Column Tableau Data Types in same order as Columns above 
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(20))    #"resource_id"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(50))    #"resource_name"
    DataSet["columns_tabdatatypes"].append(SqlType.int())          #"year"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(20))    #"color"
    DataSet["columns_tabdatatypes"].append(SqlType.varchar(20))    #"pantone_value"
    DataSet["columns_tabdatatypes"].append(SqlType.timestamp())    #"insert_date"

    fake = Faker()
    fake.add_provider(address)
    fake.add_provider(company)
    fake.add_provider(internet)
    fake.add_provider(person)
    fake.add_provider(phone_number)
    fake.add_provider(profile)

    #Define Data rows w.r.t Data parsed from SampleAPIJSONS in same order as Columns above 
    for SampleAPIJSON in SampleAPIJSONS:
        for SampleAPIData in SampleAPIJSON["data"]:
            Row = []
            Row.append(str(SampleAPIData["id"]))      #"resource_id"
            Row.append(fake.company())                               #"resource_name"
            Row.append(SampleAPIData.get("year"))                     #"year"
            Row.append(str(fake.random_number()))           #"color"
            Row.append(str(fake.random_number()))                       #"pantone_value"
            Row.append(Insert_Date)                               #"insert_date"
            DataSet["rows"].append(Row)

    #Return the DataSet
    return DataSet


and lastly....


For easy running of the Main.py script and to have the .hyper or .tdsx file uploaded to Tableau server, you could use the below Windows Batch Script. A similar thing could be achieved in UNIX as well using bash.

--> P2H_Extract_and_Publish.bat

This code should be self-explanatory and it uses tabcmd for doing the uploading to Tableau Server.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
@ECHO OFF
REM All Paths without quotes and trailing Back Slash
SET PYTHON_PATH=
SET SCRIPTS_PATH=
SET EXTRACT_PATH=
SET LOG_PATH=
SET BACKUP_PATH=

SET INCPERIOD=LastWeek

REM Comma (, ) Seperated Values for Emails Below
SET EMAIL_TO=

REM Tableau Server Details without quotes
SET TAB_SERVER=
SET TAB_PROJECT=
SET TAB_USER=
SET TAB_PASSWORD=

REM Input Parameters
SET EXTRACT_FILE_NAME=%1
SET CTMAK=%2
SET CTMSK=%3
SET LOAD_TYPE=%4

IF [%1]==[] GOTO EXIT
IF [%2]==[] GOTO EXIT
IF [%3]==[] GOTO EXIT
IF [%4]==[] GOTO EXIT


FOR /F "usebackq tokens=1,2 delims=." %%A IN (`ECHO %EXTRACT_FILE_NAME%`) DO SET FILE_NAME=%%A
FOR /F "usebackq tokens=1,2 delims=:" %%a IN (`ECHO %SCRIPTS_PATH%`) DO SET SCRIPTS_DRIVE=%%a:

IF EXIST "%EXTRACT_PATH%\%1" (
SET BACKUP=--backup yes --backuppath "%BACKUP_PATH%"
) ELSE (
SET BACKUP=--backup no
)

IF %LOAD_TYPE%==INC (
SET LOAD=--load INC --incperiod %INCPERIOD%
) ELSE (
SET LOAD=--load FULL
)

%SCRIPTS_DRIVE%
cd %SCRIPTS_PATH%


ECHO.>> %LOG_PATH%\log_schedule_%FILE_NAME%.txt
ECHO.>> %LOG_PATH%\log_schedule_%FILE_NAME%.txt

ECHO ---------------------------------------------------- >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt
ECHO Schedule Log %DATE% %TIME% >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt
ECHO ---------------------------------------------------- >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt

ECHO %PYTHON_PATH%\python.exe Main.py --ctmak %CTMAK% --ctmsk %CTMSK% --output "%EXTRACT_PATH%\%EXTRACT_FILE_NAME%" --log "%LOG_PATH%\log_%FILE_NAME%.txt" --emailto "%EMAIL_TO%" %LOAD% %BACKUP% >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt
%PYTHON_PATH%\python.exe Main.py --ctmak %CTMAK% --ctmsk %CTMSK% --output "%EXTRACT_PATH%\%EXTRACT_FILE_NAME%" --log "%LOG_PATH%\log_%FILE_NAME%.txt" --emailto "%EMAIL_TO%" %LOAD% %BACKUP% >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt

ECHO. >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt

ECHO tabcmd publish "%EXTRACT_PATH%\%EXTRACT_FILE_NAME%" --name "CTM_%FILE_NAME%_DS" --project "%TAB_PROJECT%" --overwrite --server %TAB_SERVER% --username %TAB_USER% --password %TAB_PASSWORD% >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt
tabcmd publish "%EXTRACT_PATH%\%EXTRACT_FILE_NAME%" --name "CTM_%FILE_NAME%_DS" --project "%TAB_PROJECT%" --overwrite --server %TAB_SERVER% --username %TAB_USER% --password %TAB_PASSWORD% >> %LOG_PATH%\log_schedule_%FILE_NAME%.txt

GOTO EOF

:EXIT
ECHO Invalid Input Parameters

:EOF
ECHO.>> %LOG_PATH%\log_schedule_%FILE_NAME%.txt


This script is also available inside the attached Sample Code. 


So what do you say folks? Would this be of some help, I would be happy to learn if you were able to use this somewhere in your projects and I saved some of your efforts while doing your requirement.

As always, happy coding and have a great time getting your data from Cloud APIs into Tableau and analyzing it!


No comments:

Post a Comment