Parser works incorrectly with high-precision points

[Advoskr]

Hi!

I'm new to InfluxDb and perform analysis on how to use this db engine in our project.

I currently face a very disgusting issue: when i try to write Double precision value to Db i get huge amount of columns which i didn't meant. For example, this HTTP request:

POST http://10.6.74.70:8086/write?u=root&p=root&db=netTestDb&time_precision=ms HTTP/1.1
User-Agent: InfluxDb.Net
Accept-Encoding: deflate
Content-Type: text/plain; charset=utf-8
Host: 10.6.74.70:8086
Content-Length: 808
Expect: 100-continue

MeasurementNet10 X=0.191052629701352,Y=0.852722118074411,T=0.906863938508026 101
MeasurementNet10 X=0.961525104922021,Y=0.888617412600954,T=0.770716769048347 102
MeasurementNet10 X=0.682256288678505,Y=0.796610517798276,T=0.231484161332009 103
MeasurementNet10 X=0.402903141641479,Y=0.642476850022784,T=0.600172209367236 104
MeasurementNet10 X=0.89336503292125,Y=0.553405792244433,T=0.304062585487991 105
MeasurementNet10 X=0.605175408350851,Y=0.270694734189052,T=0.632288566153631 106
MeasurementNet10 X=0.422191675017677,Y=0.435935292130306,T=0.56993827390016 107
MeasurementNet10 X=0.621569584878892,Y=0.587194737320391,T=0.350754068396405 108
MeasurementNet10 X=0.634489901659307,Y=0.0263585280749754,T=0.310414618491388 109
MeasurementNet10 X=0.651461392013105,Y=0.623513519588631,T=0.793712959528767 110

results in creating 1274 columns. Query select * from MeasurementNet10 limit 10 returns following JSON:

{  
   "results":[  
      {  
         "series":[  
            {  
               "name":"MeasurementNet10",
               "columns":[  
                  "time",
                  "0011779507627608",
                  "00164085207583422,Y",
                  "0020515057267861,Y",
                  "00533210439855797,Y",
                  "00879879622198585,Y",
                  "00959851453527739,T",
                  "00998993916902223,Y",
                  "0104352971587494,Y",
                  "0122137796190631,Y",
                  "0126514369680786,Y",
                  "0127815604269419,Y",
                  "0143174813195679,T",
                  "015039923607856",
                  "0152499703761423",
                  "016034455511735,T",
                  "0161105650552132",
                  "0180115876803229,Y",
                  "0189901413484431",
                  "0204541278166949,Y",
                  "021582078664369,Y",
                  "0222871019608747,Y",
                  "0241724108458368,Y",
                  "024625006608956,Y",
                  "0266947732431324,Y",
                  "0281805936378336,Y",
                  "0293763303334715",
                  "0301880040346589,T",
                  "0305972411439741,Y",
                  "0309916157419754,T",
                  "0310081918868274",
                  "0328753045913183",
                  "0341893718737128",
                  "0346249179144506",
                  "0353271505028602,Y",
                  "0358616574834388",
                  "0365816853179511,Y",
                  "0372902369300324,T",
                  "0374049265111819",
                  "0376476915728523,Y",
                  "0377267622564578,T",
                  "0377279524867087,Y",
                  "0384009746082132,Y",
                  "0389843383054176,T",
                  "0390534945945505,Y",
                  "0404405952619578,T",
                  "0405206610637348,T",
                  "0408780332845068,T",
                  "0419608042770814,Y",
                  "0422141622017204",
                  "0427698041511559,Y",
                  "0429770075916206,T",
                  "044194912092851",
                  "0456177252557211,Y",
                  "0465381155007231,Y",

and so on, i think you don't need full JSON to understand the problem.

So the question is: is this a bug, or a feature and what precision does Influx provide on floating point values?

[KiNgMaR]

Your line protocol syntax is incorrect, please read #2696

[Advoskr]

@KiNgMaR Can you please explain what exactly is incorrect? I follow this patter:
measurement[,tag=value,tag2=value2...] field=value[,field2=value2...] [unixnano]
Thus, i get: MeasurementNet10 for measurement, no tags, three fields with Values(X,Y,T) and time in unixnano. What do i miss?

[beckettsean]

@Advoskr you are providing a time_precision for the line protocol, which is known to lead to problems: #3044. The line protocol assumes all points are in nanosecond precision, but you provided millisecond precision in the query string.

Can you retry your writes without providing the time_precision parameter?

[Advoskr]

@beckettsean I have retried my tests without time_precision and it gave me nothing, i still face the same problem. Also, i tried less accurate point, ex. MeasurementNet10 X=0.49677,Y=0.41315,T=0.27501 12, but it also gives me this incorrect behaviour. Maybe i miss something?

[Advoskr]

@beckettsean I missed a thing. It seems that problem also corresponds to incorrect database deletion. My test performs delete database and create database before writes and it works. I also tried to drop database from web UI. Delete works, but data is incorrect (it doesn't wipe data about serie). When i used another database (newly-created) all worked fine (the data is ok). Also, there is no problem with floating point precision.
I still don't clearly understand how can i drop database correcly without harming meta or shards folders

[beckettsean]

@Advoskr thanks for the additional information. It sounds like this is another report like those in #2916 and #3049. Does that seem correct to you?

[Advoskr]

@beckettsean thanks for your responses
Yes, they look very similar to what i faced. Also it'll be great to fix a problem with time_precision in line protocol. I have read your discussion on whether or not we need precision specification when speaking about UNIX epoch, but as far as epoch can be specified in ms or even s i find it usefull to support precision feature in line protocol also. Or, at least, ignore it totally, so that we cannot make such a mistake

[beckettsean]

@Advoskr I think the write precision works fine with the line protocol, but the syntax is precision, not time_precision as appears in your initial curl example.

See the "Epoch and Precision" section on https://influxdb.com/docs/v0.9/concepts/reading_and_writing_data.html