This file lists the incompatible changes between Version 2.x and older versions

%%%%%%%%%%%%%%%%%%%%%%%%%%%% VERSION 2.2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

* General:  [ use GP function names in description ]
==========
  - Default args for user defined functions now evaluated when function is
    called (used to be: at definition time)
    E.g f(x) = local(z = x); z. 
    Before: f(2) -> x
    Now:    f(2) -> 2
  
  - rnfisnorm() input format has changed. Older version was complicated and
    only worked if rnfequation(bnf, p, 1)[3] == 0  [otherwise, crash].
    Use rnfisnorminit() now.

  - rnfidealreltoabs, rnfidealup: now return a Z-basis as a vector of
    elements in the relative extension, instead of a meaningless HNF matrix
    wrt an unknown basis. Was especially dangerous if the extension also existed
    in nfinit form, since the HNF matrices were _not_ ideals wrt this nf
    structure. [ also consistent with rnfeltreltoabs ]. Use the following
    construction instead, referring to an explicit NF structure:

      \\ return y = rnfidealup(rnf,...) as an ideal in HNF form associated to
      \\ nf = nfinit(rnf.pol);
      idealgentoHNF(nf, y) = mathnf( Mat( nfalgtobasis(nf, y) ) );

  - the "bit accuracy component" for computations of principal ideal generators
    (bnfisprincipal), units (bnfunit), and some of the bnf structures has
    been removed.

  - the "technical parameter" to bnfinit() has been shortened to 3 components
    (the others were deprecated / useless): [c, c2, nrpid]. The output format
    has changed (technical components).

  - poldegree(0) now returns -VERYBIGINT, not -1. Before using d = poldegree(x),
    always check for 0 first: either (x == 0) or (d >= 0) as you prefer. Never
    check for  d == -1 or -2^31-1 which is not portable.

  - ellheight now uses the standard normalization: twice the value it used to
    return. The values returned by ellbil() and ellheightmatrix are
    unaffected. In particular, ellheightmatrix() is the polar form of
    elleight(), and ellbil now satisfies the proper identity
    B(P, Q) = (h(P+Q)-h(P)-h(Q)) / 2

  - ':' no longer allowed as a substitute for ';' if compatible = 0. Use GP2C
    semantics [ x:int, v:vec ]. For the time being the type information is
    discarded.

  - log(x, {flag}): optional flag removed. Decide alone whether to use AGM.

  - valuation(x, 1) used to return 1, valuation(x, -1) returned 0 if (x >= 0)
    and 1 otherwise. Not anymore: error message.

  - bnrdisclist has lost its 4th argument 'flag', and omitting the
    archimedean component now means that all 2^r1 possible values are
    substituted (formerly: indicate no ramification at infinity).

  - 'buffersize' default has disappeared: it is adjusted internally whenever
    needed. The -b xxx flag to gp is a no-op.

  - the member function bnr.zkst is deprecated, use bnr.bid.clgp

  - the member function bnr.futu is deprecated. Please don't use bnr.tufu
    either.
  - use \frac instead of \over in TeX output. Define
      \def\frac#1#2{{#1\over#2}}
    if you insist on using plain TeX and run into problems.
  - automatic concatenation for strings: use longest match for expression.
    print("a"[1]) is not valid since "a" is not a vector print("a", [1])
    prints 'a[1]'.
  - isprime() now guarantees primality, use ispseudoprime() for fast
    pseudo-primality tests. 
  - default() now always returns the value of the (possibly changed) default.
    No need for a flag anymore.
  - semantic of t_SER with inexact coefficients is now the same as for
    t_POL: the sign is 0 iff all coefficients are zero. Either there are
    no coefficients, or the leading coefficient is an inexact zero.
  - No longer assume that part of an object is "permanent" when it is out
    of the stack (was used by INTMOD/POLMOD/PADIC). Always copy it.
    As a result, 'gmodulo' and 'forcecopy' become obsolete.
    Rename gmodulcp -> gmodulo.

* Specific to the PARI library:
===============================
Incompatible changes:
---------------------
  - gsize() renamed to sizedigit()  [conflict with gtk]
  - hnfhavas removed (didn't work properly, hnflll provides an alternative)
  - real zeroes are now coded on 2 words. Beware of constructs like
    t = cgetg(lg(x), t_REAL); gaffect(y, t). If x = 0, so will be t.
  - polgalois(): old format deprecated. New preferred format for result 
    has 3rd component giving numbering among all transitive subgroups of S_n
    [ was ad hoc up to 7, as described above for n >= 8 ]. Old format is
    still the default, but will eventually change. Use
    default(new_galois_format, 1) or, in library mode, set global variable
    new_galois_format to 1 to enable the new format.
  - The nf structure output by nfinit has changed: 
    *) nf.zk is now T2-LLL-reduced, not in HNF wrt the power basis
    *) the internal components of nf[5] have changed (MC and T2 not needed
       anymore)
  - polred & polredabs do not take a 'prec' argument anymore [was unused]
  - gentimer / genmsgtimer / get_timer have been removed. They are superseded
    by TIMER and msgTIMER which are fully reentrant and easier to use.
  - (undocumented) macros buch[gen | genfu | init | iniftu]
  - rnfisnorm() prototype has changed [ + need to call rnfisnorminit first ]
  - incgam4 renamed to incgam0, incgam3 renamed to incgamc
  - prototypes of buchall and smallbuchinit have changed (much simpler)
  - co8() renamed to quadtoc()
  - ker_mod_p() renamed to FpM_ker [ and supplemented with many analogous
    modular routines ]
  - nfreducemodpr2() removed, use an nfmodpr structure from nfmodprinit() or
    zkmodprinit() instead
  - smodsi() removed [ not well defined ]. Use either modsi() or umodui()
  - Many error codes were removed. It is usually better to use only talker
    outside libpari.           
  - Macro BITS_IN_RANDOM has been removed. Used to be 32.
  - mymyrand() has been removed, use pari_rand31() or pari_rand() instead
  - diviiz(x,y,z), divisz, divsiz and divssz always assign the euclidean
    quotient [ used to  depend on the type of z: if t_REAL computed exact
    quotient ]. Use rdivii, rdivis, rdivsi, rdivss for analogous
    functionality (no "z" variant);
  - mpdivz(x,y,z) assigns the euclidean quotient when x,y are t_INT (used
    to depend on the type of z)
  - removed inconsistently named macros mpinv[sir]r [ were "z" functions ]
  - removed useless routine shifts [ use shifti( stoi() ) ]
  - library interface of functions intnum, prodeuler, suminf, sumalt,
       sumalt2, sumpos, sumpos2, prodinf, prodinf1  [ use GEN
       (*eval)(GEN,void*) everywhere instead of entree * ]
  - changed the prototype of bnrdisclist0
  - removed ideallistarch0, ideallistarchgen, ideallistunitarch,
    ideallistunitarchgen. Just use ideallistarch.
  - change the output of ideallist with technical (flag 2,3) [ with units ]:
    instead of two vectors, output a vector of 2-component vectors.
    change the input of all list routines (bnrclassnolist, bnrdisclist)
    accordingly.
  - rename gtrans_i -> shallowtrans, concatsp -> shallowconcat.
  - pari_err(warner | warnmem | warnfile | warnprec,) no longer accepted. Use
    pari_warn.

Partially compatible changes:
-----------------------------
[ not mandatory for your program to work with 2.3, since compatibility macros
support the old names but are likely to become mandatory in 2.4, so we
advise you to update soon ]

  - typecasting macros (e.g. ladd(), lmul() ...) are obsolete and
    shouldn't be used in new programs. Use accessors gel() and friends.
  - stack locations used to be of type long or ulong, now they have a
    dedicated type pari_sp [ pari stack pointer ]
  - in order to compare variable numbers, use the macro varncmp() instead of
    < or > operators. 
  - access to the prime numbers table is done via the macros            
    NEXT_PRIME_VIADIFF or NEXT_PRIME_VIADIFF_CHECK.   
  - direct access to the mantissa of t_INT is deprecated. Instead you should   
    use the t_INT API (macros int_MSW, int_LSW, int_precW, int_nextW, int_W).
    This will ensure your code is compatible with both native and GMP kernels.
  - the macros lgef / setlgef / evallgef have been removed. t_POLs no longer
    include an "effective length". One should use lg() for t_POLs as for
    most other types. Don't use lgef() in new code.

    lgef() and setlgef() are aliased to lg() and setlg(). They will break on
    hackish code accessing directly the first component of t_POL objets, e.g
      x = cgetg(20, t_POL);
      x[1] = evalsigne(1) | evalvarn(0) | evallgef( 10 )  /* 10, not 20 */
      ...
      setlgef(x, 20); /* now create/use further coefficients */
  - types t_FRACN, t_RFRACN and gred() have been removed (complicated, unused,
    and very inefficient). There are now aliases to t_FRAC, t_RFRAC and gcopy
    respectively for backward compatibility.
  - renamed nfdivres -> nfdivrem, poldivres -> poldivrem
  - renamed permute -> numtoperm, permuteInv -> pertonum
  - renamed gzero -> gen_0, gun -> gen_1, gdeux -> gen_2, polx -> pol_x,
    polun -> pol_1
  - renamed gegal -> gequal, gegalgs -> gequalgs, gegalsg -> gequalsg,
    egalii -> equalii
  - renamed binome -> binomial, chinois -> chinese
  - renamed apprgen9, apprgen -> padicappr, factmod9 -> factorff
  - rename resss -> remss, ressi -> remsi, resis -> remis, resii -> remii,
    gres -> grem
  - rename divise -> dvdii, gdivise -> gdvd, mpdivis -> dvdiiz, mpdivisis
    -> dvdisz
  - rename mpent -> mpfloor
  - rename isprincipalrayall -> bnrisprincipal, rayclassno -> bnrclassno
    rayclassnolist -> bnrclassnolist
  - rename globalreduction -> ellglobalred, localreduction -> elllocalred
  - taniyama(e) is deprecated. Use elltaniyama(e, prec) instead of
      old = precdl; precdl = prec; x = taniyama(e); precdl = old;
  - rename lisexpr -> readexp, lisseq -> readseq, flisexpr -> freadexpr,
    flisseq -> freadseq
  - rename idmat -> matid
  - rename coefs_to_col -> mkcoln, coefs_to_int -> mkintn
           coefs_to_pol -> mkpoln, coefs_to_vec -> mkvecn
  - rename wf -> weberf, wf1 -> weberf1, wf2 -> weberf2
  - rename err -> pari_err, pariputsf -> pariprintf
  - rename rnfhermitebasis -> rnfhnfbasis
  - rename gcarreparfait -> gissquare, gcarrecomplet -> gissquarerem
  - rename lseriesell -> elllseries
  - gmodulo and forcecopy are obsolete. Use gmodulcp and gcopy.

%%%%%%%%%%%%%%%%%%%%%%%%%%%% VERSION 2.1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

* The calculator GP:
====================
 - most function names have changed (see misc/new.dic or use whatnow under GP)
 - lowercase / uppercase letters distinguished.
 - the syntax \var = value is not recognized anymore. Use standard
   metacommands instead.
(setting the compatibility level "compatible" to 3, with "default" will give
you those three back).

 - strict parenthesis matching before executing the command
(set strictmatch to 0 to get old behaviour)

 - functions label / goto have been removed (use break/next/return instead).
 - conjugation x_ has been removed. '_' can be freely used in identifiers.
 - \k metacommand removed

 - result history is now cyclic (older results are erased when the maximal
   size "histsize" is reached).
 - the Set() function turns objects into strings (so that set functions are
   able to treat arbitrary objects). Use eval to turn them back to GENs.
 - the type() function returns a string and not a number.
 - sum(X=a,b,...,x) yields x if b<a. Analogous statement for prod.

* The library PARI:
===================
 - use symbolic type names (t_INT, etc.). Don't use any explicit type number
   in your programs since these are liable to change (they have not, but will).
 - the codewords of GEN objects have a different internal structure. The
   components should only be accessed through the documented macros. When
   creating codeword x[1] for the GEN x, use the eval* macros (the set* ones
   don't fill the codeword completely).
 - the following constant names have changed ("pari" prepended): C1, C2, C3, K,
   K1, K2, K4. The constant C31 has disappeared.
 - prec and defaultpadicprecision are no longer global. The latter has
   disapeared.
 - char *pariversion not set anymore (use the macros PARIVERSION/PARIINFO)
 - LONG_IS_32BIT is never defined (test whether LONG_IS_64BIT is defined
   or not).
 - gen2str() renamed to GENtostr().
 - gitoascii() suppressed, use GENtostr().
 - init() renamed to pari_init().
 - the macro mant() has been suppressed.
 - imprimer() suppressed. output() is now a function and can be used under
   all debuggers.
 - the prototypes and behaviour of gredsp() and normalize() have changed.
   They are not documented anymore.
 - the install() function has been modified (use codes instead of valence).
 - gpuigs only has two arguments.
 - integ now takes two arguments, like deriv.
 - setrand, getrand, getstack, gettime, glength return a C long, and not a GEN.
 - pari_randseed no longer global, use setrand/getrand
 - gerepile(ltop,lbot,0) no longer returns ltop - lbot
 - nf[5][7] stored in two element form (was Z-basis) for faster inversion
 - t_POL and t_POLMOD can be freely mixed as long as main variables are the
   same (result is the expected POLMOD)