Data loss on big encoded polyline

[ismail BASKIN]

it is true as expected

--it is true as expected
SELECT ST_IsClosed(ST_LineFromEncodedPolyline('guhiFwmpqEs]fJwmAbmAcmAngAwGve@wGrcBg|@~bBo_@bQ{c@fTos@vLs]bGscBsD{uAkf@cuB{aAkMsD{kAc[os@rDs`AzOs{@zTwj@zOkCnZoPntBoPf|@oKzOseAj}A{c@zw@c[bj@s]bmAcQvj@oAfEcVfw@s{@vvDce@~gBwQfr@_b@rv@s]ve@s{@nU{}B~C{c@kCsl@gEkCSo|CsDkHoAk_A_SklBw`@s`AgJchA?kH?_eAoUon@{Os]sIkiASs`Ab[cpBf|@_`Ajf@wBz@kMvGgYvrAbBz^rDn}@jf@znBbLrhBjCjwDg@bBsD~McBnF{Jf^wVr{@{EfO{zAbuBcG~H{zA~bB{Of|@_b@jxAos@f|@od@jH_q@bLowCnZkbB?kk@Ss`Ab[s`A~bBwhA~oCco@f{CgOb`@kMb[{JvL{JvLkRnUcrAzOwmAgJ_`AzT_b@zaAwj@~p@gYfw@wGbrAbBrq@rDrhBvGf|@nAjnArNf^fw@zEjf@nF~f@zErv@v~@jz@rgEvV~eCfOrkCvLrIsDj`HrNzY~p@bwAjRb`@zJ{@jW_DvrAgOfEoAbcAgY~a@g^jCcGjWco@nPw[zOoZrXg^jH{JfYs]zJkMbo@sb@rcBkbBnZ{YfYgObGSzm@oA~oCbhAj_An_@faF~iArI~Cve@bQrrBbt@bQnUjRvVbLzO~sArtAnx@nU~a@be@z^be@ni@fh@rXfc@jWvVjRvo@zTfw@~Wnx@fc@v[be@bVvVzJb~@j\jp@R~Hzc@rIja@z@rIjCrSrDbj@rD~f@zE~a@vBb[vBj\bBz^nAnZSz^?~\Srb@wBnn@oArXcBni@kCf^?fc@SnFcBb`@g@~f@_Dbe@_Drg@kCn_@{Ej\_Drb@kCfYwBfJz^fmEr]~sFzTrgE?z_G??jC~oCbLnUvj@rDbrArDraCfJjpEsD~bBjH??zYz@zxBfEbpBzOjlBnUr~Ave@jlBvrAjiAz|@nvAjsAzr@nlAf|@~p@vj@fJ~_Ab[zh@r]z^vVjHzEzxBjxAznBbcAju@n_@f}BngA~gBnUve@jf@vmAngAbQnZ~\~k@r~AfJzw@sDvmAkk@nn@kk@fYcmAjCkxAkMseAwBcQcGkf@oPscBzT{fAbL_l@vj@gOzY{@r{@wB~_AgJbj@?~C?nqA_l@jH_DjbBogA~iAnUr{@fJ~nAr]nFbBvmAsD~M_DnbAoUjf@_l@zw@_q@vhA{T~dAgJzzAfJr~A~k@~dA~p@vmArq@fEz@vy@~Rbo@_Dbo@w`@?kk@oPw`@co@{Ts`A{OcmAkf@{EkH_q@wcAwGkxAbL{sBvj@gnBvhAkxAzE{@zc@kHns@g|@jk@kRbGoAbcAgfAf@g@ns@ogAfYogA?cmAwGclDcLwdCwGgnBoPcmA?_~AkHg^oKgh@_b@_~As]sIgYf@gxBv`@kdAbt@oyBbwAgc@Rw[{JoZgY_`AkeCos@okDk_AkkEkM{m@cBcGkp@gbCs]suC?scBbLcrAbo@_~Azw@shBvQkf@zh@stAjCwGjCcmAkf@czBgw@_cB_eAczB_eAgnBwL_l@gT{aA{TwdCkCwdC?_bERsDnK_~Ar]ogAns@g|@zEwBjsAsg@jlB_q@rg@_g@nZs]nU_X~\seArIgh@kHgwEwGkk@_b@{aAcLsq@{w@kxAcLogAka@g|@{^cj@w`@gm@{^os@k\co@oPg|@wBwLcQwy@wGwrARgEvBo~BkCwrA{TgnB_Ik\sS_{@?we@oPsq@sXc`@os@wrAg_Bwy@od@oUsg@_X{zAsv@crA{|@gw@scB_l@ovA{JwV{aAsrBkCwGgfAczBwt@c|AS_NbGos@g@cVwG_b@oP_SkRgTg_Bc[s~A_l@s`AgJwj@rDco@rv@g^j_AsSfh@g|@rq@_eAS{ToZ{TwrAoKkf@s]cmA{JcQcVod@wpB?gOkHkk@kWgYkk@oPsq@'));
 

it is false, because ST_AsEncodedPolyline-ST_LineFromEncodedPolyline corrupt data.

--it is false, because ST_AsEncodedPolyline-ST_LineFromEncodedPolyline corrupt data.
SELECT ST_IsClosed(ST_LineFromEncodedPolyline(ST_AsEncodedPolyline(ST_LineFromEncodedPolyline('guhiFwmpqEs]fJwmAbmAcmAngAwGve@wGrcBg|@~bBo_@bQ{c@fTos@vLs]bGscBsD{uAkf@cuB{aAkMsD{kAc[os@rDs`AzOs{@zTwj@zOkCnZoPntBoPf|@oKzOseAj}A{c@zw@c[bj@s]bmAcQvj@oAfEcVfw@s{@vvDce@~gBwQfr@_b@rv@s]ve@s{@nU{}B~C{c@kCsl@gEkCSo|CsDkHoAk_A_SklBw`@s`AgJchA?kH?_eAoUon@{Os]sIkiASs`Ab[cpBf|@_`Ajf@wBz@kMvGgYvrAbBz^rDn}@jf@znBbLrhBjCjwDg@bBsD~McBnF{Jf^wVr{@{EfO{zAbuBcG~H{zA~bB{Of|@_b@jxAos@f|@od@jH_q@bLowCnZkbB?kk@Ss`Ab[s`A~bBwhA~oCco@f{CgOb`@kMb[{JvL{JvLkRnUcrAzOwmAgJ_`AzT_b@zaAwj@~p@gYfw@wGbrAbBrq@rDrhBvGf|@nAjnArNf^fw@zEjf@nF~f@zErv@v~@jz@rgEvV~eCfOrkCvLrIsDj`HrNzY~p@bwAjRb`@zJ{@jW_DvrAgOfEoAbcAgY~a@g^jCcGjWco@nPw[zOoZrXg^jH{JfYs]zJkMbo@sb@rcBkbBnZ{YfYgObGSzm@oA~oCbhAj_An_@faF~iArI~Cve@bQrrBbt@bQnUjRvVbLzO~sArtAnx@nU~a@be@z^be@ni@fh@rXfc@jWvVjRvo@zTfw@~Wnx@fc@v[be@bVvVzJb~@j\jp@R~Hzc@rIja@z@rIjCrSrDbj@rD~f@zE~a@vBb[vBj\bBz^nAnZSz^?~\Srb@wBnn@oArXcBni@kCf^?fc@SnFcBb`@g@~f@_Dbe@_Drg@kCn_@{Ej\_Drb@kCfYwBfJz^fmEr]~sFzTrgE?z_G??jC~oCbLnUvj@rDbrArDraCfJjpEsD~bBjH??zYz@zxBfEbpBzOjlBnUr~Ave@jlBvrAjiAz|@nvAjsAzr@nlAf|@~p@vj@fJ~_Ab[zh@r]z^vVjHzEzxBjxAznBbcAju@n_@f}BngA~gBnUve@jf@vmAngAbQnZ~\~k@r~AfJzw@sDvmAkk@nn@kk@fYcmAjCkxAkMseAwBcQcGkf@oPscBzT{fAbL_l@vj@gOzY{@r{@wB~_AgJbj@?~C?nqA_l@jH_DjbBogA~iAnUr{@fJ~nAr]nFbBvmAsD~M_DnbAoUjf@_l@zw@_q@vhA{T~dAgJzzAfJr~A~k@~dA~p@vmArq@fEz@vy@~Rbo@_Dbo@w`@?kk@oPw`@co@{Ts`A{OcmAkf@{EkH_q@wcAwGkxAbL{sBvj@gnBvhAkxAzE{@zc@kHns@g|@jk@kRbGoAbcAgfAf@g@ns@ogAfYogA?cmAwGclDcLwdCwGgnBoPcmA?_~AkHg^oKgh@_b@_~As]sIgYf@gxBv`@kdAbt@oyBbwAgc@Rw[{JoZgY_`AkeCos@okDk_AkkEkM{m@cBcGkp@gbCs]suC?scBbLcrAbo@_~Azw@shBvQkf@zh@stAjCwGjCcmAkf@czBgw@_cB_eAczB_eAgnBwL_l@gT{aA{TwdCkCwdC?_bERsDnK_~Ar]ogAns@g|@zEwBjsAsg@jlB_q@rg@_g@nZs]nU_X~\seArIgh@kHgwEwGkk@_b@{aAcLsq@{w@kxAcLogAka@g|@{^cj@w`@gm@{^os@k\co@oPg|@wBwLcQwy@wGwrARgEvBo~BkCwrA{TgnB_Ik\sS_{@?we@oPsq@sXc`@os@wrAg_Bwy@od@oUsg@_X{zAsv@crA{|@gw@scB_l@ovA{JwV{aAsrBkCwGgfAczBwt@c|AS_NbGos@g@cVwG_b@oP_SkRgTg_Bc[s~A_l@s`AgJwj@rDco@rv@g^j_AsSfh@g|@rq@_eAS{ToZ{TwrAoKkf@s]cmA{JcQcVod@wpB?gOkHkk@kWgYkk@oPsq@'))));

[woodbri]

The encode and decode function take a precision=5 argument. This means that it truncates the floating point numbers at that number of significant decimal point digits. So if you encode and decode a linestring you will not get back "exactly" the same linestring. This is the definition of the encoding and decoding functions. So there is no error here other then understanding what the function is doing.

[ismail BASKIN]

Example polyline encoded on 5 precision (34.4958 38.3882,34.494 38.3931...). So these functions should not truncate values.

[woodbri]

Thanks for example, this will help someone narrow down the issue.

[ismail BASKIN]

Is there any improvement?

[robe]

No. Still get same answer in PostGIS 2.4 dev. @dbaston any thoughts on this. I vaguely remember a limitation of representing things like polygons so wondering if that affects closed rings as well though why LineFrom wouldn't suffer the same issue doesn't make sense.

[robe]

I'm guessing this won't be done anytime soon. Push back if you can get to it in next day or 2.

[pramsey]

Here's an example that is small enough to debug:

with d as (select 'SRID=4326;LINESTRING(33.6729 38.7071,33.6692 38.701,33.6673 38.6972,33.6626 38.6871)'::geometry as g),
e as (
   select g, ST_LineFromEncodedPolyline(ST_AsEncodedPolyline(g)) as gl
   from d
   ) 
select 
  st_astext(g), ST_AsEncodedPolyline(g), st_npoints(g), 
  st_astext(gl), st_npoints(gl)
from e

yuck.

[pramsey]

In 16130:

Support google line encodings that include a backslash character (References #3713)

[pramsey]

In 16131:

Support google line encodings that include a backslash character (References #3713)

[pramsey]

In 16132:

Support google line encodings that include a backslash character (References #3713)

[pramsey]

There were actually two things going on. The original author thought that it would be smart to escape the backslash character in the output, but actually the format didn't like that, it preferred the raw character. Also the postgis/ side caller was prematurely freeing the database memory underneath, which caused failures *sometimes*. Fix applied to trunk, 2.4, 2.3.